CN103280218A - Selection method based on voice recognition, mobile terminal device and information system thereof - Google Patents

Abstract

A selection method based on voice recognition, a mobile terminal device and an information system thereof are provided. The selection method comprises the following steps: receiving a first voice input; performing voice recognition and natural language processing on the first voice input to generate a corresponding first keyword; obtaining at least one first reply answer according to the first keyword; when the number of the first return answers is 1, performing corresponding operation according to the type corresponding to the first return answer; when the number of the first return answers is more than 1, sorting the first return answers, displaying the first return answers in a first candidate list and receiving second input voice; performing voice recognition and natural language processing on the second input voice to generate a corresponding second keyword; and selecting a second return answer from the second return answers in the first candidate list according to the second keyword.

Description

Selection method based on voice recognition, mobile terminal device and information system thereof

technical field

The invention relates to a selection method, a mobile terminal device and an information system thereof, in particular to a voice recognition-based selection method, a mobile terminal device and an information system. the

Background technique

In the computer's natural language understanding (Nature Language Understanding), a specific grammar is usually used to capture the intention or information of the user's input sentence. Therefore, if there are enough data of user input sentences stored in the database, a reasonable judgment can be made. the

In the existing practice, one is to use the built-in fixed word list to capture the user's input sentence, and the fixed word list contains specific intentions or specific terms used in information, and the user needs to follow this specific term. Only by expressing its intention or information can its intention or information be correctly recognized by the system. However, forcing the user to memorize each specific term of the fixed word list is quite inhumane. For example: the prior art uses the implementation mode of fixed word list, requires the user to say when asking weather: "Shanghai (or Beijing) tomorrow (or the day after tomorrow) weather how?" Also want to ask about weather conditions, such as "how is Shanghai tomorrow?", because "weather" does not appear in the sentence, so the prior art will be understood as "there is a place called tomorrow in Shanghai", which obviously does not catch the user's true intent. In addition, the types of sentences used by users are very complex and change from time to time. Sometimes users may even input wrong sentences. In this case, it is necessary to capture the user's input sentences by means of fuzzy matching. Therefore, it is even less effective to provide only fixed word lists with rigid input rules. the

In addition, when natural language understanding is used to process multiple types of user intentions, the grammatical structure of some different intentions is the same. For example, when the user's input sentence is "I want to watch Romance of the Three Kingdoms", the user intention may be You want to watch a movie about Romance of the Three Kingdoms, or you want to read a book about Romance of the Three Kingdoms, so usually in this case, two possible intentions will be matched for the user to choose. However, in many cases, it is superfluous and inefficient to provide unnecessary possible intents for users to choose. For example, when the user's input sentence is "I want to watch the Super Avenue of Fame", it is very unnecessary to match the user's intention as reading a book or painting on the Super Avenue of Fame (because the Super Avenue of Fame is a TV program). the

Furthermore, generally speaking, the search results obtained in the full-text search are unstructured data. The information in unstructured data is scattered and unrelated. For example, after entering keywords in search engines such as Google or Baidu, the webpage search results obtained are unstructured data, because the search results must be read item by item manually. In order to find the useful information in it, and such an approach not only wastes the user's time, but also may miss the desired information, so the practicality will be greatly limited. the

Contents of the invention

The invention provides a selection method based on voice recognition, a mobile terminal device and an information system thereof, which can improve the convenience of user operation. the

The present invention proposes a selection method based on speech recognition, including: receiving a first speech input; performing speech recognition on the first speech input to generate a first keyword; generating at least one first return answer according to the first keyword; when selecting When the number of first reported answers is 1, perform the corresponding operation according to the data type of the selected first reported answer; when the number of selected first reported answers is greater than 1, display the first candidate list containing the first reported answer and receive second voice input; perform voice recognition on the second voice input to generate a second keyword; select the second reported answer from the first reported answers displayed in the first candidate list according to the second keyword. the

The invention proposes a mobile terminal device, which includes a voice receiving unit, a display unit, a storage unit and a data processing unit. The voice receiving unit receives the first voice input and the second voice input. The display unit is used for displaying a candidate list including the reported answer. The storage unit is used for storing a plurality of data. The data processing unit is coupled to the voice receiving unit, the display unit and the storage unit. The data processing unit conducts voice recognition on the first voice input to generate a first keyword, and selects a corresponding first reported answer according to the first keyword. When the number of selected first reported answers is 1, the data processing unit performs a corresponding operation according to the type of the selected first reported answer. When the number of selected first reported answers is greater than 1, the data processing unit controls the display unit to display the first candidate list including the first reported answers. The data processing unit conducts voice recognition on the second voice input to generate a second keyword, and selects the second reported answer from the first reported answers in the first candidate list according to the second keyword. the

The invention proposes an information system, including a server and a mobile terminal device. The server is used for storing a plurality of data and has a voice recognition function. The mobile terminal device includes a voice receiving unit, a display unit and a data processing unit. The voice receiving unit receives the first voice input and the second voice input. The display unit is used to display a candidate list containing the returned answer. The data processing unit is coupled to the voice receiving unit, the display unit and the server. The data processing unit performs voice recognition on the first voice input through the server to generate a first keyword, and the server selects a corresponding first reported answer according to the first keyword and sends it to the data processing unit. When the number of selected first reported answers is 1, the data processing unit performs a corresponding operation according to the type of the selected first reported answer. When the number of selected first reported answers is greater than 1, the data processing unit controls the display unit to display the first candidate list containing the first reported answer, and the data processing unit performs voice recognition on the second voice input through the server to generate the second keywords, and the server selects a second reported answer from the first reported answers in the first candidate list according to the second keyword and sends it to the data processing unit. the

The present invention proposes a selection method based on speech recognition, including: searching a structured database according to the first keyword to obtain at least one first return answer; when the number of the selected first return answer is greater than 1, displaying a first candidate data including the first reported answer; receiving a second voice input after displaying the first candidate list, and performing voice recognition on the second voice input to generate a second keyword; and, according to The second user intends to select a second reported answer from the first reported answer in the first candidate list. the

Based on the above, the speech recognition-based selection method and its mobile terminal device and information system in the embodiment of the present invention perform speech recognition and natural language processing on the first speech input and the second speech input to confirm the first speech input and the second speech input. The corresponding keyword is voice input, and the reported answer is selected according to the keywords corresponding to the first voice input and the second voice input. Thereby, the convenience of the user's operation can be improved. the

In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail with reference to the accompanying drawings. the

Description of drawings

FIG. 1 is a block diagram of a natural language understanding system according to an embodiment of the present invention. the

Fig. 2 is a schematic diagram of analysis results of various request information of users by a natural language processor according to an embodiment of the present invention. the

FIG. 3A is a schematic diagram of a plurality of records with a specific data structure stored in a structured database according to an embodiment of the present invention. the

FIG. 3B is a schematic diagram of a plurality of records with a specific data structure stored in a structured database according to another embodiment of the present invention. the

FIG. 3C is a schematic diagram of guidance data stored in the guidance data storage device according to an embodiment of the present invention. the

FIG. 4A is a flowchart of a retrieval method according to an embodiment of the present invention. the

FIG. 4B is a flow chart of the working process of the natural language understanding system according to another embodiment of the present invention. the

FIG. 5A is a block diagram of a natural language dialogue system according to an embodiment of the present invention. the

FIG. 5B is a block diagram of a natural language understanding system according to an embodiment of the present invention. the

FIG. 5C is a block diagram of a natural language dialogue system according to another embodiment of the present invention. the

FIG. 6 is a flowchart of a method for modifying a voice response according to an embodiment of the invention. the

FIG. 7A is a block diagram of a natural language dialogue system according to an embodiment of the present invention. the

FIG. 7B is a block diagram of a natural language dialogue system according to another embodiment of the present invention. the

FIG. 8A is a flowchart of a natural language dialogue method according to an embodiment of the present invention. the

FIG. 8B is a schematic diagram of multiple records with specific data structures stored in the structured database according to yet another embodiment of the present invention. the

FIG. 9 is a system diagram of a mobile terminal device according to an embodiment of the present invention. the

FIG. 10 is a system diagram of an information system according to an embodiment of the present invention. the

FIG. 11 is a flowchart of a selection method based on voice recognition according to an embodiment of the present invention. the

FIG. 12 is a block diagram of a voice control system according to an embodiment of the present invention. the

FIG. 13 is a block diagram of a voice control system according to another embodiment of the present invention. the

FIG. 14 is a flowchart of a voice control method according to an embodiment of the present invention. the

Description of reference symbols

100, 520, 520’, 720, 720’: Natural Language Understanding System

102, 503, 503’, 703, 902, 902’: request information

104: Analysis Results

106: Possible intent syntax data

108, 509, 509', 711, 904, 904': keywords

110: Response result

112: Intent Data

114: Determine intent syntax data

116: Analysis result output module

200: retrieval system

220: Structured Databases

240: Search Engines

260: Retrieve interface unit

280: Guide Data Storage Device

300: Natural Language Processor

302, 832, 834, 836, 838: record

304: Title field

306: Content field

308: subfield

310: guide field

312: Numeric field

314: source field

316: Heat field

318, 852, 854: preferences field

320, 862, 864: disgust field

400: Knowledge Assisted Comprehension Module

500, 500’, 700, 700’: Natural Language Dialogue System

501, 701: voice input

507, 507', 707: voice response

510, 710: voice sampling module

511, 511', 711, 906, 906': return the answer

513, 513', 713: Speech

522, 722: Speech recognition module

524, 724: Natural Language Processing Module

526, 726: speech synthesis module

530, 740: speech synthesis database

702: Speech integrated processing module

715: User preference data

717: User preferences record

730: Feature Database

872, 874: fields

900, 1010: mobile terminal device

908, 908': candidate list

910, 1011: voice receiving unit

920, 1013: data processing unit

930, 1015: display unit

940: storage unit

1000: Information Systems

1020: Server

SP1: First Voice

SP2: Second Speech

1200, 1300: voice control system

1210: Auxiliary starting device

1212, 1222: wireless transmission module

1214: trigger module

1216: Wireless charging battery

12162: battery unit

12164: Wireless charging module

1220, 1320: mobile terminal device

1221: Voice system

1224: Speech sampling module

1226: Speech Synthesis Module

1227: Voice output interface

1228: Communication module

1230: (Cloud) Server

1232: Speech Understanding Module

12322: Voice recognition module

12324: Voice processing module

S410～S450: Steps of the retrieval method according to an embodiment of the present invention

S510～S590: Steps in the working process of the natural language understanding system according to an embodiment of the present invention

S602, S604, S606, S608, S610, S612: Each step of the method for correcting the voice response

S802～S890: each step of the natural language dialogue method according to an embodiment of the present invention

S1100～S1190: each step of the selection method based on speech recognition according to an embodiment of the present invention

S1402～S1412: each step of the voice control method according to an embodiment of the present invention

Detailed ways

Because the existing implementation of using the fixed word list can only provide rigid input rules, the judgment ability for the user's changeable input sentence is very insufficient, so it often leads to wrong judgment of the user's intention and the required information cannot be found, or It is a problem such as outputting unnecessary information to the user due to insufficient judgment. In addition, existing search engines can only provide users with scattered and weakly relevant search results, so users have to spend time viewing items one by one to filter out the required information, which not only wastes time but also may miss the required information. The present invention proposes a structured data retrieval method and system for the foregoing problems of the prior art, and provides specific fields in the structured data to store different types of data elements, so that when users use natural voice input information for retrieval, they can Quickly and accurately judge the user's intention, and then provide the required information to the user, or provide more accurate information for them to choose. the

FIG. 1 is a block diagram of a natural language understanding system according to an embodiment of the present invention. As shown in Figure 1, the natural language understanding system 100 includes a retrieval system 200, a natural language processor 300 and a knowledge-assisted understanding module 400, the knowledge-assisted understanding module 400 is coupled to the natural language processor 300 and the retrieval system 200, and the retrieval system 200 also includes The structured database 220 , the search engine 240 and the search interface unit 260 , wherein the search engine 240 is coupled to the structured database 220 and the search interface unit 260 . In this embodiment, the retrieval system 200 includes a retrieval interface unit 260, but it is not intended to limit the present invention. In some embodiments, the retrieval interface unit 260 may not be present, and in other ways the search engine 240 performs a full-text search on the structured database 220. search. the

When the user sends request information 102 to the natural language understanding system 100, the natural language processor 300 can analyze the request information 102, and send the analyzed possible intention grammar data 106 to the knowledge-aided understanding module 400, wherein the possible intention grammar data 106 Contains keywords 108 and intent data 112 . Subsequently, the knowledge-assisted understanding module 400 takes out the keywords 108 in the possible intent grammar data 106 and sends them to the retrieval system 200 and stores the intent data 112 inside the knowledge-assisted understanding module 400, and the search engine 240 in the retrieval system 200 will be based on the keywords After the word 108 performs full-text search on the structured database 220 , the response result 110 of the full-text search is sent back to the knowledge-aided understanding module 400 . Next, the knowledge-aided understanding module 400 compares the stored intent data 112 according to the response result 110, and sends the determined intent grammar data 114 obtained to the analysis result output module 116, and the analysis result output module 116 then determines the The intent grammar data 114, transmits the analysis result 104 to the server (not shown), and then sends it to the user after querying the data required by the user. It should be noted that the analysis result 104 may include the keyword 108 , and may also output partial information (such as the number of the record 302 ) or all information of the record (such as the record in FIG. 3A/3B ) containing the keyword 108 . In addition, the analysis result 104 can be directly converted into voice by the server and output to the user, or the corresponding voice can be output to the user after specific processing (the method of "specific processing" and the content and information included will be described in detail later. ), those skilled in the art can design the information output by the retrieval system 200 according to actual needs, which is not limited in the present invention. the

The above-mentioned analysis result output module 116 can be combined with other modules as appropriate, for example, in one embodiment, it can be incorporated into the knowledge-aided understanding module 400, or in another embodiment, it can be separated from the natural language understanding system 100 and located in the server In the server (for example, including the natural language understanding system 100), the server will directly receive the intent grammar data 114 for processing. In addition, the knowledge-aided understanding module 400 can store the intent data 112 in a storage device inside the module, in the natural language understanding system 100, in a server (such as including the natural language understanding system 100), or in any available knowledge The auxiliary comprehension module 400 can retrieve the storage, which is not limited in the present invention. Furthermore, the natural language understanding system 100 including the retrieval system 200, the natural language processor 300, and the knowledge-aided understanding module 400 can be constructed with hardware, software, firmware, or various combinations of the above methods, and the present invention does not cover this Limit. the

The aforementioned natural language understanding system 100 can be located in a cloud server, a server in a local area network, or even a personal computer, a mobile computer device (such as a notebook computer), or a mobile communication device (such as a mobile phone). The components of the natural language understanding system 100 or the retrieval system 200 do not necessarily need to be installed in the same machine, but may be distributed in different devices or systems and connected by various communication protocols according to actual needs. For example, the natural language understanding processor 300 and the knowledge-assisted understanding module 400 can be configured in the same smart phone, and the retrieval system 200 can be configured in another cloud server; or, the retrieval interface unit 260, the natural language understanding processor 300 and the knowledge aided comprehension module 400 can be configured in the same notebook computer, and the search engine 240 and the structured database 220 can be configured in another server in the local area network. In addition, when the natural language understanding system 100 is located on a server (whether it is a cloud server or a local network server), the retrieval system 200, the natural language understanding processor 300, and the knowledge-aided understanding module 400 can be configured on different computers Host, and the main server system coordinates the transmission of messages and data between them. Certainly, the retrieval system 200 , the natural language understanding processor 300 , and the knowledge-aided understanding module 400 can also combine two or all of them in a computer host according to actual needs, and the present invention does not limit the configuration of this part. the

In the embodiment of the present invention, the user can send request information to the natural language processor 300 in various ways, for example, send the request information by speaking voice input or text description. For example, if the natural language understanding system 100 is located in the cloud or in a server (not shown) in a local area network, the user can first input the request through a mobile device (such as a mobile phone, PDA, tablet computer or similar system) Information 102, and then the request information 102 is sent to the natural language understanding system 100 in the server through the telecommunications system operator, so that the natural language processor 300 can analyze the request information 102, and finally the server confirms the user's intention, and then After the corresponding analysis result 104 is processed by the server through the analysis result output module 116 , the information requested by the user is sent back to the user's mobile device. For example, the request information 102 may be a question to which the user wishes to obtain an answer through the natural language understanding system 100 (such as "how is the weather in Shanghai tomorrow"), and the natural language understanding system 100 analyzes the user's intention is When inquiring about the weather in Shanghai tomorrow, the inquired weather data will be sent to the user as the output result 104 through the analysis result output module 116 . In addition, if the user's instruction to the natural language understanding system 100 is "I want to watch Let the Bullets Fly" or "I want to listen to the days we walked together", because "Let the Bullets Fly" or "The Days We Walked Together" may contain different fields, so the natural language processor 300 will analyze the user's request information 102 into one or more possible intent grammar data 106, and the possible intent grammar data 106 includes keywords 108 and intent data 112, and then After performing a full-text search on the structured data 240 in the search system 220, the user's intention is further confirmed. the

Further, when the user's request information 102 is "how is the weather in Shanghai tomorrow, after analysis by the natural language processor 300, a possible intent grammar data 106 can be generated:

"<queryweather>,<city>=Shanghai,<time>=tomorrow". the

In one embodiment, if the natural language understanding system 100 thinks that the user's intention is quite clear, it can directly output the user's intention (that is, query the weather in Shanghai tomorrow) through the analysis result output module 116 to output the analysis result 104 to the server, The server can send the weather information specified by the user to the user after querying it. For another example, when the user's request information 102 is "I want to watch the Romance of the Three Kingdoms", the natural language processor 300 can produce three possible intention grammar data 106 after analysis:

"<readbook>,<bookname>=Romance of the Three Kingdoms"; 

"<watchTV>,<TVname>=Romance of the Three Kingdoms"; and

"<watchfilm>,<filmname>=Romance of the Three Kingdoms". the

This is because the keywords 108 (i.e. "Romance of the Three Kingdoms") in the syntax data 106 may belong to different fields, i.e. books (<readbook>), TV dramas (<watchTV>), and movies (<readfilm>) Three domains, so one request information 102 can be analyzed into multiple possible intent grammar data 106, so it needs to be further analyzed by the knowledge-assisted comprehension module 400 to confirm the user's intent. To give another example, if the user enters "I want to watch Let the Bullets Fly", because "Let the Bullets Fly" may be the title of a movie or a book, at least the following two possible intent grammars may appear Data 106:

"<readbook>,<bookname>=let the bullets fly"; and

"<watchfilm>,<filmname>=let the bullets fly"; 

They belong to two fields of books and movies respectively. The above-mentioned possible intent grammar data 106 needs to be further analyzed by the knowledge-assisted comprehension module 400, and the definite intent grammar data 114 is obtained therefrom to express the clear intention of the user's requested information. When the knowledge aided comprehension module 400 analyzes the possible intent grammar data 106 , the knowledge aided comprehension module 400 can transmit the keywords 108 (such as the aforementioned “Romance of the Three Kingdoms” or “Let the bullets fly”) to the retrieval system 200 through the retrieval interface 260 . The structured database 220 in the retrieval system 200 stores multiple records with a specific data structure, and the search engine 240 can perform a full-text search on the structured database 220 through the keyword 108 received by the retrieval interface unit 260, and retrieve the full-text The response result 110 obtained by the search is sent back to the knowledge-aided understanding module 400 , and then the knowledge-aided understanding module 400 can use the response result 110 to obtain the definite intent grammar data 114 . As for the details of performing a full-text search on the structured database 220 to determine the intent grammar data 114 , it will be described in more detail later through FIG. 3A , FIG. 3B and related paragraphs. the

In the concept of the present invention, the natural language understanding system 100 can first capture the keywords 108 in the user's request information 102, and use the full-text search results of the structured database 220 to determine the domain attributes of the keywords 108, such as the above-mentioned input When "I want to watch Romance of the Three Kingdoms", possible intention grammar data 106 belonging to the three fields of books, TV dramas and movies will be generated, and then the user's clear intention will be further analyzed and confirmed. Therefore, users can easily express their intentions or information in a colloquial manner without needing to memorize specific terms, such as the specific terms related to the fixed word list in the existing practice. the

FIG. 2 is a schematic diagram of analysis results of various request information of a user by a natural language processor 300 according to an embodiment of the present invention. the

As shown in Figure 2, when the user's request information 102 is "how is the weather in Shanghai tomorrow", after analysis by the natural language processor 300, the possible intent grammar data 106 can be generated as follows:

"<queryweather>,<city>=Shanghai,<time>=tomorrow"

The intent data 112 is "<queryweather>", and the keywords 108 are "Shanghai" and "tomorrow". Since only one set of intent grammar data 106 (query weather <queryweather>) is obtained after analysis by the natural language processor 300, in one embodiment, the knowledge-aided understanding module 400 can directly extract the keywords 108 "Shanghai" and "tomorrow". "As the analysis result 104, it is sent to the server to query weather information (for example, to query tomorrow's Shanghai weather overview, including information such as weather, temperature, etc.), without the need to perform a full-text search on the structured database 220 to determine the user's intention. Of course, in one embodiment, the full-text search of the structured database 220 can still be performed to determine the user's intention more accurately, and those skilled in the art can make changes according to actual needs. the

In addition, when the user's request information 102 is "I want to watch and let the bullets fly", because two possible intent syntax data 106 can be produced:

"<readbook>,<bookname>=let the bullets fly"; and

"<watchfilm>,<filmname>=let the bullets fly"; 

Two corresponding intent data 112 "<readbook>" and "<watchfilm>" and two identical keywords 108 "let the bullets fly" indicate that the intent may be to read books or "let the bullets fly" Watch the "Let the Bullets Fly" movie. In order to further confirm the user's intention, the keyword 108 "let the bullets fly" will be sent to the retrieval interface unit 260 through the knowledge-aided understanding module 400, and then the search engine 240 will use the keyword 108 "let the bullets fly" to search the structured database 220 A full-text search is performed to confirm that "let the bullets fly" should be the title of a book or movie to confirm the user's intent. the

Furthermore, when the user's request information 102 is "I want to listen to the days I walked together", two possible intent grammar data 106 can be produced:

"<playmusic>,<singer>=walked together,<songname>=days"; "<playmusic>,<singer>=days walked together" 

Two corresponding identical intent data 112 "<playmusic>" and two sets of corresponding keywords 108 "walked together" and "days" and "days walked together", to indicate that the intention may be to listen to the singer The song "days" sung by "walking together", or listening to the song "days walking together", at this time, the knowledge aided understanding module 400 can transmit the first group of keywords 108 "walking together" and "days" and The second group of keywords "the days we walked together" is given to the retrieval interface unit 260 to confirm whether there is the song "days" sung by the singer of "walking together" (the implied user of the first group of keywords intent), or whether there is the song "The Days We Walked Together" (the implied user intent of the second group of keywords), so as to confirm the user's intent. However, the present invention is not limited to the formats and names corresponding to the possible intent grammar data and intent data represented here. the

FIG. 3A is a schematic diagram of a plurality of records with a specific data structure stored in the structured database 220 according to an embodiment of the present invention. the

Generally speaking, in some existing full-text search methods, the obtained search results are unstructured data (such as the results searched by Google or Baidu), because the information of the search results is scattered and irrelevant , so the user has to view each piece of information one by one, thus causing practical limitations. However, in the concept of the present invention, the efficiency and accuracy of retrieval can be effectively improved by using a structured database. Because the numerical data contained in each record in the structured database disclosed by the present invention are related to each other, and these numerical data are jointly used to express the attributes of the record. Therefore, when the search engine performs a full-text search on the structured database, when the recorded numerical data matches the keyword, it can output the guide data corresponding to the numerical data as confirmation of the intention of the request information. The implementation details of this part will be further described through the following examples. the

In an embodiment of the present invention, each record 302 stored in the structured database 220 includes a title field 304 and a content field 306, the title field 304 includes a plurality of subfields 308, and each subfield includes a guide field 310 and a value field 312 , the guide field 310 of the plurality of records 302 is used to store guide data, and the value field 312 of the plurality of records 302 is used to store value data. Take the record 1 shown in Figure 3A as an example to illustrate here, three sub-fields 308 in the title field 304 of record 1 store respectively:

"singerguid: Andy Lau",

"songnameguid: the days we walked together"; and

"songtypeguid: Hong Kong and Taiwan, Cantonese, popular"; 

The guidance field 310 of each subfield 308 stores the guidance data "singerguid", "songnameguid" and "songtypeguid" respectively, and the numerical field 312 corresponding to the subfield 308 stores the numerical data "Andy Lau", "Walking together Days" and "Hong Kong and Taiwan, Cantonese, Popular". The guidance data "singerguid" represents the field type of the numerical data "Andy Lau" is the singer name (singer), the guidance data "songnameguid" represents the field type of the numerical data "The days we walked together" is the song name (song), and the guidance data "songtypeguid" The field type of "representative numerical data" Hong Kong and Taiwan, Cantonese, popular" is song type. In fact, each guide data here can be represented by a different specific series of numbers or characters, which is not limited in the present invention. The content field 306 of record 1 is to store the lyrics content of the song "The Days We Walked Together" or store other data (such as composer/lyricist... etc.), but the real data in the content field 306 of each record is not The emphasis of the invention is therefore only schematically depicted in FIG. 3A . the

In the aforementioned embodiments, each record includes a title field 304 and a content field 306, and the sub-field 308 in the title field 304 includes a guide field 310 and a value field 312, but this is not intended to limit the present invention, and in some embodiments it can also be There is no content field 306 and, in some embodiments, there may be no directions field 310 . the

In addition, in the embodiment of the present invention, a first special character is stored between the data of each sub-field 308 to separate the data of each sub-field 308, and between the data of the index field 310 and the value field 312 is stored a A second special character to separate the data in the index field and the value field. For example, as shown in Figure 3A, between "singerguid" and "Andy Lau", between "songnameguid" and "the days we walked together", and between "songtypeguid" and "Hong Kong and Taiwan, Cantonese, popular" are The second special character ":" is used to separate, and the sub-fields 308 of record 1 are separated by the first special character "|", but the present invention is not limited to ":" or "|" is the special character used to separate. the

On the other hand, in an embodiment of the present invention, each subfield 308 in the title field 304 can have a fixed number of digits, for example, the fixed number of digits of each subfield 308 can be 32 characters, and the fixed number of the guide field 310 can be 32 characters. The number of digits can be 7 or 8 digits (used to guide 128 or 256 different guide data at most). In addition, because the required digits of the first special character and the second special character can be fixed, the sub-field 308 After deducting the fixed number of digits occupied by the guide field 310 , the first special character, and the second special character, the remaining number of digits can be used to store the numerical data of the value field 312 . Furthermore, since the number of digits in the sub-field 308 is fixed, plus the content of the data stored in the sub-field 308, as shown in FIG. data, the second special character, and as previously mentioned, the number of bits of these four data is also fixed, so in practice, the bits of the index field 310 can be skipped (such as skipping the first 7 or 8 bits), and After the number of digits of the second special character (such as skipping 1 character, that is, 8 digits), and then deducting the digits occupied by the first special character (such as the last character, 8 digits), finally Just can directly obtain the numerical data of numerical field 312 (for example, directly take numerical data "Andy Lau" in the first sub-field 308 of record 1, now there are 32-3=29 characters available for storing the numerical value of numerical field 312 data, 3 (that is, 1+1+1) in the formula represents 1 character occupied by the guidance data, the first special character and the second special character of the directed field 310 respectively), and then proceed to the required field The type can be judged. Thus, after the comparison of the numerical data taken out at present is completed (regardless of whether the comparison is successful or not), the numerical data of the next sub-field 308 can be taken out according to the above-mentioned method of taking out the numerical data (for example, in the second field of record 1). directly take out the numerical data "the days we walked together" in the sub-field 308) to compare the types of comparison fields. The above method of taking out numerical data can be compared from record 1, and after comparing all the numerical data in record 1, take out the numerical data of the first sub-field 308 in the title field 308 of record 2 (such as "Feng Xiaogang ”) for comparison. The above comparison procedure will continue until all recorded numerical data have been compared. the

It should be noted that the above-mentioned number of digits in the sub-field 308 , and the number of digits used in the index field 310 , the first special character, and the second special character may vary according to actual applications, and the present invention is not limited thereto. The above method of extracting numerical data by comparison is just an example, but it is not intended to limit the present invention. Another example can be performed by using a full-text search method. In addition, the implementation of the above-mentioned skip guide field 310, the second special character, and the first special character can be achieved by bit shifting (such as division), and the implementation of this part can be implemented in hardware, software, or a combination of both Those skilled in the art can make changes according to actual needs. In another embodiment of the present invention, each sub-field 308 in the title field 304 may have a fixed number of digits, the index field 310 in the sub-field 308 may have another fixed number of digits, and the title field 304 may not include the first For special characters and second special characters, since the digits of each subfield 308 and each guide field 310 are fixed, each subfield can be directly taken out by skipping a certain number of digits or by bit shifting (such as division). Guidance data or numeric data in field 308. the

It should be noted that, as mentioned above, the sub-field 308 has a certain number of digits, so a counter can be used in the natural language understanding system 100 (or in a server including the natural language understanding system 100) to record the currently compared It is a certain field 308 of a certain record. In addition, the compared records can also use another counter to store their sequence. For example, when a first counter record is used to represent the current sequence of compared records, and a second counter is used to represent the current sequence of sub-fields compared, if the current comparison is the record in Figure 3A During the 3rd sub-field 308 of 2 (that is, when comparing "filenameguid: Huayi Brothers"), the stored value of the first counter will be 2 (representing that the current comparison is record 2), and the stored value of the second counter is The value is then 3 (indicating that the current comparison is the third sub-field 308). Furthermore, the above-mentioned method of only storing the guidance data of the guidance field 310 with 7 or 8 bits is to hope that most of the words in the sub-field 308 are used to store numerical data, and the actual guidance data can be passed through these 7 or 8 bits. The bit is used as an indicator, and then read the actual guidance data from the guidance data storage device 280 stored in the retrieval system 200, wherein the guidance data is stored in the form of a table, but any other information that can be accessed by the retrieval system 200 All methods can be used in the present invention. Therefore, in actual operation, in addition to directly taking out the numerical data for comparison, it is also possible to directly take out the guide data as the response result 110 and send it to the knowledge-aided understanding module based on the values of the above two counters when generating the matching result 400. For example, when the second sub-field 308 of record 6 (i.e. "songnameguid: betrayal") is successfully matched, it will be known that the current first counter/second counter values are 6 and 2 respectively, so it can be based on These two values are stored in the guide data storage device 280 shown in FIG. 3C , and the guide data is “songnameguid” found in sub-field 2 of record 6 . In one embodiment, after the number of digits of the sub-field 308 can be fixed, all the bits of the sub-field 308 can be used to store numerical data, so the index field, the first special character, and the second special character can be completely removed, and The search engine 240 only needs to know that it is another sub-field 308 every time the fixed number is crossed, and add one to the second counter (of course, the storage value of the first counter needs to be increased by one every time the next record is retrieved) , which provides more bits to store numeric data. the

Another example is given to illustrate the process of returning the matching record 110 to the knowledge-aided understanding module 400 for further processing when the comparison generates a matching result. Corresponding to the data structure of the above-mentioned record 302, in an embodiment of the present invention, when the user's request information 102 is "I want to watch let the bullets fly", two possible intent syntax data 106 can be produced:

"<readbook>,<bookname>=let the bullets fly"; with

"<watchfilm>,<filmname>=let the bullets fly"; 

The search engine 240 performs a full-text search on the title field 304 of the records stored in the structured database 220 of FIG. 3A by using the keyword 108 "let the bullets fly" received by the search interface unit 260 . In the full-text search, the record 5 storing the value data "let the bullets fly" is found in the title field 304, thus a matching result is generated. Next, the retrieval system 200 returns the guidance data “filmnameguid” corresponding to the keyword 108 “let the bullets fly” in the title field 304 of the returned record 5 as the response result 110 and returns it to the knowledge-aided understanding module 400 . Since the title field of record 5 contains the guidance data "filmnameguid" corresponding to the numerical data "let the bullets fly", the knowledge aided understanding module 400 compares the guidance data "filmnameguid" of record 5 with the above-mentioned possible intent grammar data 106 previously Stored intent data 112 "<watchfilm>" or "<readbook>", it can be judged that the definite intent syntax data 114 of this request information is "<watchfilm>, <filmname>=let bullets fly" (because they all contain "film" in it). In other words, the data "Let the Bullets Fly" described in the user's request information 102 this time is the title of a movie, and the data user's request information 102 intends to watch the movie "Let the Bullets Fly" instead of reading books. the

Give another example for further explanation. When the user's request information 102 is "I want to listen to the days I walked through together", two possible intent grammar data 106 can be produced:

"<playmusic>,<singer>=walked together,<songname>=days"; with

"<playmusic>,<songname>=the days we walked together"; 

The search engine 240 then retrieves the two groups of keywords 108 received by the interface unit 260:

"walked together" and "days"; and

"The days we walked together"

To perform a full-text search on the title field 304 of the records stored in the structured database 220 of FIG. 3A . Because in the full-text search, the matching results corresponding to the first group of keywords 108 "walking together" and "day" were not found in the title field 304 of all records, but the matching results corresponding to the second group of keywords 108 " The record 1 of the days we walked together", so the retrieval system 200 returns the guide data "songnameguid" corresponding to the second group of keywords 108 in the title field 304 of the record 1 as the matching record 110 and returns it to the knowledge-aided understanding module 400. Next, after receiving the guidance data "songnameguid" corresponding to the numerical data "days walked together", the knowledge aided comprehension module 400 walks together with the possible intent grammar data 106 (that is, "<playmusic>,<singer>= ,<songname>=days" and "<playmusic>, <songname>=days you walked together") in the intent data 112 (that is, <singer>, <songname>, etc.) are compared, so it is found that this time The user's request information 102 does not describe the data with the name of the singer, but the data with the title of the song "the days we walked together" (because only <songname> is compared successfully). Therefore, the knowledge-assisted comprehension module 400 can judge that the definite intention grammar data 114 of the request information 102 is "<playmusic>, <songname>=the days we walked together" through the above comparison, and the user's request information 102 The intention is to listen to the song "The Days We Walked Together". the

In another embodiment of the present invention, the retrieved response result 110 may be an exact match record that completely matches the keyword 108 , or a partial match record that partially matches the keyword 108 . For example, if the user's request information 102 is "I want to hear Hsiao Jingteng's betrayal", similarly, after analysis, the natural language processor 300 produces two possible intention grammar data 106:

"<playmusic>,<singer>=Xiao Jingteng,<songname>=betrayal"; and "<playmusic>,<singer>=Xiao Jingteng's betrayal"; 

And send two sets of keywords 108: 

"Xiao Jingteng" and "Betrayal"; and

"Xiao Jingteng's Betrayal";

To the search interface unit 260, the search engine 240 then uses the keyword 108 received by the search interface unit 260 to perform a full-text search on the title field 304 of the record 302 stored in the structured database 220 of FIG. 3A. Because in the full-text search, no records were matched for the second group of keywords 108 "Xiao Jingteng's betrayal", but the matching results of records 6 and 7 were found for the first group of keywords 108 "Xiao Jingteng" and "betrayal" . Since the second group of keywords 108 "Xiao Jingteng" and "betrayal" only matched the numerical data "Xiao Jingteng" in record 6, but did not match other numerical data "Yang Zongwei" and "Cao Ge", record 6 is a partial match record (Please note that the above record 5 corresponding to the request information 102 "I want to watch let the bullets fly" and the record 1 corresponding to the request information "I want to listen to the days we walked together" are both partial matching records), and the keyword "Xiao Jingteng" and "Betrayal" completely matches the numerical data of record 7 (because the second group of keywords 108 "Xiao Jingteng" and "betrayal" both match successfully), so record 7 is an exact match record. In the embodiment of the present invention, when the retrieval interface unit 260 outputs a plurality of matching records 110 to the knowledge-aided understanding module 400, it can output the complete matching records (that is, all the numerical data are matched) and the partial matching records in sequence (that is, only part of the numerical data is matched) matching records 110, wherein the priority of the complete matching records is higher than that of the partial matching records. Therefore, when the search interface unit 260 outputs the matching record 110 between record 6 and record 7, the output priority of record 7 will be higher than that of record 6, because all the numerical data "Xiao Jingteng" and "betrayal" of record 7 are generated. Matching results, but record 6 also contains "Yang Zongwei" and "Cao Ge" and did not produce matching results. That is to say, the higher the matching degree of the records stored in the structured database 220 to the keywords 108 in the requested information 102 , the easier it is to be preferentially output, so that the user can check or select the corresponding definite intent grammar data 114 . In another embodiment, the matching record 110 corresponding to the record with the highest priority may be directly output for use in determining the intent grammar data 114 . The foregoing is not intended to limit the present invention, because in another embodiment, it may be possible to output as long as a matching record is found (for example, taking "I want to hear Xiao Jingteng's betrayal" as the request information 102, when the record 6 That is, when the matching result is generated, the guide data corresponding to the output record 6 is used as the matching record 110 ), without prioritization, so as to speed up the retrieval. In another embodiment, the record with the highest priority can be directly processed and provided to the user. For example, when the highest priority is to play the movie Romance of the Three Kingdoms, the movie and the user can be played directly. In addition, if the highest priority is Betrayal sung by Xiao Jingteng, this song can be played directly to the user. It should be noted that the present invention is only described here and not limited thereto. the

In yet another embodiment of the present invention, if the user's request information 102 is "I want to listen to Andy Lau's betrayal", one of the possible intended grammar data 106 is:

"<playmusic>,<singer>=Andy Lau,<songname>=betrayal"; 

If the search interface unit 260 inputs the keywords 108 "Andy Lau" and "betrayal" into the search engine 240, no matching result will be found in the database in FIG. 3A. In yet another embodiment of the present invention, the search interface unit 260 can input the keywords 108 "Andy Lau" and "betrayal" into the search engine 240, and obtain correspondingly whether "Andy Lau" is the name of the singer (guidance data singerguid) and " Betrayal" is the name of the song (guide data songnameguid, and the singer may be Cao Ge or Xiao Jingteng, Yang Zongwei and Cao Ge). At this time, the natural language understanding system 100 can further remind the user: "Is the song Betrayal sung by Xiao Jingteng (according to the matching result of record 7)?" matching results)?". the

In yet another embodiment of the present invention, the records stored in the structured database 220 may further include a source field 314 and a popularity field 316 . As shown in FIG. 3B , in addition to the fields in FIG. 3A , it also includes a source field 314 , popularity field 316 , like field 318 and dislike field. The source field 314 of each record can be used to store which structured database the record comes from (only the structured database 220 is shown in this diagram, but in fact there may be more different structured databases), or which user , the origin value provided by the server. Moreover, the natural language understanding system 100 can search a structured database of a specific source according to the preference disclosed by the user in the previous request message 102 (for example, when a full-text search is performed with the keyword 108 in the request message 102 to generate a match, then the The heat value of the record is incremented by one). And the popularity field 316 of each record 302 is used to store the search popularity value or the popularity degree value of this record 302 (for example, this record is matched by a single user, a specific user group, or the number of times or probability of all users in a specific time), so as to For reference when the knowledge-assisted comprehension module 400 judges the user's intention, the use of the like field 318 and the dislike field will be described in detail later. In detail, when the user's request information 102 is "I want to watch Romance of the Three Kingdoms", after analysis by the natural language processor 300, multiple possible intent grammar data 106 can be generated:

"<readbook>,<bookname>=Romance of the Three Kingdoms"; 

"<watchTV>,<TVname>=Romance of the Three Kingdoms"; and

"<watchfilm>,<filmname>=Romance of the Three Kingdoms". the

If the natural language understanding system 100 counts most of the requests for watching movies in the historical records of the user's request information 102 (for example, using the heat field 316 to store the number of times the record 302 is clicked by a certain user), Then the natural language understanding system 100 can search the structured database for storing movie records (at this time, the source value in the source field 314 is the code for recording the structured database for storing movie records), so that it can preferentially determine "<watchfilm> ,<filmname>=Romance of the Three Kingdoms" is the grammatical data 114 for determining the intent. For example, in one embodiment, each record 302 can be matched once, and one can be added to the popularity field 316 as the user's historical record. Therefore, when performing a full-text search based on the keyword 108 "Romance of the Three Kingdoms", the record 302 with the highest value in the popularity field 316 can be selected from all matching results for judging the user's intention. In one embodiment, if the natural language understanding system 100 determines that the search popularity value stored in the popularity field 316 of the TV program corresponding to the TV program "Romance of the Three Kingdoms" is the highest among the search results of the keyword 108 "Romance of the Three Kingdoms", then Then it can be preferentially determined that "<watchTV>, <TVname>=Romance of the Three Kingdoms" is the grammatical data 114 for determining the intention. In addition, the above-mentioned way of changing the value stored in the popularity field 316 can be changed through the computer system where the natural language understanding system 100 is located, which is not limited in the present invention. In addition, the value of the popularity field 316 can also decrease with time to indicate that the user's popularity for a certain record 302 has gradually decreased, and the present invention does not limit this part. the

To give another example, in another embodiment, because the user may particularly like to watch the TV series of Romance of the Three Kingdoms in a certain period of time, because the length of the TV series may be very long and the user cannot watch it in a short time, so in a short time It may be repeatedly clicked (assuming that the value in the popularity field 316 is increased by one every time it is matched), thus causing a certain record 302 to be matched repeatedly, which can be known by analyzing the data in the popularity field 316. Moreover, in another embodiment, the telecommunications company can also use the popularity field 316 to indicate the popularity of the data provided by a certain source, and the code of the data provider can be stored in the source field 314 . For example, if a supplier who supplies "Romance of the Three Kingdoms" has the highest probability of being selected, so when a user inputs the request information 102 of "I want to watch Romance of the Three Kingdoms", although the database in Fig. 3B is processed During the full-text search, three matching results will be found: reading Romance of the Three Kingdoms (Record 8), watching Romance of the Three Kingdoms TV series (Record 9), and watching Romance of the Three Kingdoms movies (Record 10). It is the most popular option now (that is, the values of the heat fields of records 8, 9, and 10 are 2, 5, and 8 respectively), so the guide data of record 10 will be provided first to make matching record 110 and output to the knowledge-assisted understanding system 400, As the highest priority option to determine user intent. In one embodiment, the data in the source field 314 can be displayed to the user at the same time, allowing the user to judge whether the TV series he wants to watch is provided by a certain provider. It should be noted that the above-mentioned data stored in the source field 314 and its modification method can also be modified through the computer system where the natural language understanding system 100 is located, which is not limited by the present invention. It should be noted that those skilled in the art should know that the information stored in the popularity field 316, favorite field 318, and dislike field 320 in FIG. 3B can be further divided into two parts related to the individual user and related to all users. And the popularity field 316, liking field 318, and dislike field 320 information related to the user will be stored in the user's mobile phone, and the server then stores information such as the popularity field 316, liking field 318, and dislike field 320 related to all users. In this way, the personal preference information related to the user's personal choice or intention is only stored in the user's personal mobile communication device (such as a mobile phone, tablet computer, or small laptop, etc.), while the server stores Information related to all users, which not only saves the storage space of the server, but also keeps the privacy of the user's personal preferences. the

Obviously, the numerical data contained in each record in the structured database disclosed by the present invention are related to each other (for example, the numerical data "Andy Lau" in record 1, "days we walked together", "Hong Kong and Taiwan , Cantonese, and popular” are used to describe the characteristics of record 1), and these numerical data are used to express the intention of the record in the request information from the user (for example, when a matching result is generated for “the days we walked together”, it means The user's intention may be to access the data of record 1), so when the search engine performs a full-text search on the structured database, when the numerical data of the record is matched, it can output the guide data corresponding to the numerical data (for example, output " songnameguid" as the response result 110), and then confirm the intent of the request information (for example, compare in the knowledge-aided understanding module 400). the

Based on the contents disclosed or taught in the above exemplary embodiments, FIG. 4A is a flowchart of a retrieval method according to an embodiment of the present invention. Please refer to Fig. 4A, the retrieval method of the embodiment of the present invention comprises the following steps:

Provide a structured database, and the structured database stores multiple records (step S410);

Receive at least one keyword (step S420);

A full-text search is performed on the title fields of multiple records by keywords (step S430). For example, the keyword 108 is input into the search interface unit 260 to allow the search engine 240 to perform a full-text search on the title fields 304 of the plurality of records 302 stored in the structured database 220. As for the search method, it can be as shown in FIG. 3A or FIG. 3B the manner in which the search is carried out, or is carried out without altering its spirit;

It is judged whether there is a matching result in the full-text search (step S440). For example, use the search engine 240 to determine whether the full-text search corresponding to the keyword 108 has matching results; and

If there is a matching result, a full matching record and a partial matching record are sequentially output (step S450). For example, if there is a record matching the keyword 108 in the structured database 220, the retrieval interface unit 260 sequentially outputs the guide data in the complete matching record and the partial matching record matching the keyword 108 (can be obtained by referring to FIG. 3C (obtained by guiding the data storage device 280) is sent to the knowledge-aided understanding system 400 as the response result 110, wherein the priority of the complete matching record is higher than that of the partial matching record. the

On the other hand, if there is no matching result, you can directly notify the user that the matching failed and end the process, notify the user that no matching result was found and ask for further input, or list possible options for the user to make further choices (such as the aforementioned " "Andy Lau" and "betrayal" do a full-text search without producing a matching result) (step 460). the

The foregoing process steps are not intended to limit the present invention, and some steps can be ignored or removed. For example, in another embodiment of the present invention, a matching judgment module (not shown in the figure) outside the retrieval system 200 can be used to ) to execute step S440; or in another embodiment of the present invention, the above-mentioned step S450 can be ignored, and the action of sequentially outputting complete matching records and partial matching records can be performed by a matching result output module located outside the retrieval system 200 (not shown in the figure), to execute the action of sequentially outputting the complete matching record and the partial matching record in step S450. the

Based on the content disclosed or taught in the above exemplary embodiments, FIG. 4B is a flow chart of the working process of the natural language understanding system 100 according to another embodiment of the present invention. Referring to Fig. 4B, the working process of the natural language understanding system 100 of another embodiment of the present invention includes the following steps:

Receive request information (step S510). For example, the user sends the request information 102 with voice content or text content to the natural language understanding system 100;

Provide a structured database, and the structured database stores multiple records (step S520);

Syntaxize the request information (step S530). For example, after the natural language processor 300 analyzes the user's request information 102, it is converted into the corresponding possible intent grammar data 106;

Identify possible attributes of keywords (step S540). For example, the knowledge-assisted comprehension module 400 identifies possible attributes of at least one keyword 108 in the grammar data 106, for example, the keyword 108 "Romance of the Three Kingdoms" may be books, movies and TV programs;

Perform a full-text search on the title fields 304 of multiple records by using the keyword 108 (step S550 ). For example, the keyword 108 is input into the search interface unit 260 to allow the search engine 240 to perform a full-text search on the title field 304 of a plurality of records stored in the structured database 220;

It is judged whether there is a matching result in the full-text search (step S560). For example, use the search engine 240 to determine whether the full-text search corresponding to the keyword 108 has matching results;

If there is a matching result, the guidance data corresponding to the complete matching record and the partial matching record (step S570 ) are sequentially output as the response result 110 . For example, if there is a record matching the keyword 108 in the structured database 220, the retrieval interface unit 260 sequentially outputs the guidance data corresponding to the complete matching record and the partial matching record matching the keyword 108 as the response result 110,

where exact match records take precedence over partial match records; and

Sequentially output the corresponding determined intent syntax data (step S580). For example, the knowledge-assisted comprehension module 400 outputs the corresponding determined intent grammar data 114 by sequentially outputting the complete matching record and the partial matching record. the

On the other hand, if no matching result is generated in step S560, it can also be processed in a manner similar to step S460, such as directly notifying the user of matching failure and ending the process, notifying the user that no matching result is found and requiring further input, or Enumerate possible options for the user to make further selections (for example, the aforementioned example of "Andy Lau" and "betrayal" full-text search did not produce matching results) (step S590). the

The foregoing process steps are not intended to limit the present invention, and some steps may be omitted or removed. the

To sum up, the present invention extracts the keywords included in the user's request information, and performs a full-text search for the title field of the record with a specific data structure in the structured database. If a matching result is generated, it can be judged that The category of the field to which the keyword belongs, so as to determine the intention expressed by the user in requesting information. the

Next, we will explain more about the application of the above structured database in speech recognition. First, in the natural language dialogue system, the application of correcting the wrong voice response according to the user's voice input, and further finding out other possible answers to report to the user is described. the

As mentioned above, although today's mobile communication devices can provide a natural language dialogue function to allow users to communicate with the mobile communication device by making a voice. However, in the current voice dialogue system, when the user's voice input is unclear, the system may easily output a voice response that does not conform to the voice input because the same voice input may mean multiple different intentions or purposes. Therefore, in many dialogue situations, it is difficult for users to get voice responses that meet their intentions. For this reason, the present invention proposes a method for correcting a voice response and a natural language dialogue system, wherein the natural language dialogue system can correct the wrong voice response according to the user's voice input, and further find out other possible answers to give back to user. In order to make the content of the present invention clearer, the following specific examples are given as examples in which the present invention can actually be implemented. the

FIG. 5A is a block diagram of a natural language dialogue system according to an embodiment of the present invention. Please refer to FIG. 5A , the natural language dialogue system 500 includes a speech sampling module 510 , a natural language understanding system 520 , and a speech synthesis database 530 . In one embodiment, the voice sampling module 510 is used to receive the first voice input 501 (such as voice from the user), and then analyze it to generate the first request information 503, and the natural language understanding system 520 will then analyze the first The request information 503 is parsed to obtain the first keyword 509 therein, and after finding the first return answer 511 that meets the first request information 503 (according to the description in Figure 1, the first request information 503 can use the same as the request information 102 In other words, the request information 102 will generate possible intent grammar data 106 after analysis, and the keywords 108 in it will be used to perform full-text search on the structured database 220 to obtain the response result 110, and the response result 110 will be compared with The intent data 112 in the possible intent grammar data 106 is compared to generate the definite intent grammar data 114, and finally the analysis result 104 is sent by the analysis result output module 116, and the analysis result 104 can be used as the first reported answer 511 in FIG. 5A), According to the first reported answer 511, a corresponding speech query is performed on the speech synthesis database 530 (because the analysis result 104 as the first answer 511 may include the relevant data of the complete/partially matched record 302 (such as stored in the guide field 310) guide data, the numerical data in the numerical field 312, and the data in the content field 306... etc.), so these data can be used for voice query), and then the first voice 513 of the query is output to generate the corresponding first voice input 501 The first voice response 507 is given to the user. Wherein, if the user thinks that the first voice response 507 output by the natural language understanding system 520 does not conform to the first request information 503 in the first voice input 501, the user will input another voice input, such as the second voice input 501', to indicate this. The natural language understanding system 520 will process the second voice input 501' in the same manner as the first voice input 501 above to generate the second request information 503', and then analyze the second request information 503' to obtain the first request information 503'. Two keywords 509', find the second reported answer 511' that matches the second request information 503', find out the corresponding second voice 513', and finally generate the corresponding second voice response 507' output according to the second voice 513' For the user, it is used for correcting the answer 511 of the first report. Obviously, the natural language understanding system 520 can be based on the natural language understanding system 100 in FIG. 1, and add a new module (will be explained in conjunction with the subsequent FIG. 5B) to achieve the purpose of correcting the wrong voice response according to the user's voice input. Purpose. the

The various components in the aforementioned natural language dialogue system 500 can be configured in the same machine. For example, the voice sampling module 510 and the natural language understanding system 520 are configured in the same electronic device. Among them, the electronic device can be a mobile phone (Cell phone), a personal digital assistant (Personal Digital Assistant, PDA) mobile phone, a smart phone (Smart phone) and other mobile communication devices, a palmtop computer (Pocket PC), a tablet computer (Tablet PC), notebook computer, personal computer, or other electronic devices with communication functions or installed with communication software, without limiting the scope here. In addition, the above-mentioned electronic device may use an Android operating system, a Microsoft operating system, an Android operating system, a Linux operating system, etc., but is not limited thereto. Of course, the various components in the aforementioned natural language dialogue system 500 do not necessarily need to be installed in the same machine, but can be distributed in different devices or systems and connected through various communication protocols. For example, the natural language understanding system 520 can be located in a cloud server, or a server in a local area network. In addition, the various components in the natural language understanding system 520 may also be distributed in different machines, for example, the various components in the natural language understanding system 520 may be located in the same or different machines as the speech sampling module 510 . the

In this embodiment, the voice sampling module 510 is used to receive voice input. The voice sampling module 510 can be a device for receiving audio such as a microphone (Microphone), and the first voice input 501/second voice input 501' can be from the user voice. the

In addition, the natural language understanding system 520 of this embodiment can be implemented by a hardware circuit composed of one or several logic gates. Alternatively, in another embodiment of the present invention, the natural language understanding system 520 may be implemented by computer program codes. For example, the natural language understanding system 520 is implemented in application programs, operating systems, or drivers, etc., by program code segments written in a programming language, and these program code segments are stored in the storage unit and processed by the processing unit (not shown in Figure 5A) to perform. In order for those skilled in the art to further understand the natural language understanding system 520 of this embodiment, examples are given below for illustration. However, the present invention is only illustrated here and is not limited thereto. For example, hardware, software, firmware, or a combination of these three implementations can be used to implement the present invention. the

FIG. 5B is a block diagram of a natural language understanding system 520 according to an embodiment of the present invention. Referring to FIG. 5B , the natural language understanding system 520 of this embodiment may include a speech recognition module 522 , a natural language processing module 524 and a speech synthesis module 526 . Wherein, the voice recognition module 522 will receive the request information transmitted from the voice sampling module 510, such as the first request information 503 for analyzing the first voice input 501, and take out one or more first keywords 509 (such as FIG. 1A keywords 108 or phrases, etc.). The natural language processing module 524 can analyze these first keywords 509 again, and obtain a candidate list containing at least one return answer (the processing method is the same as that of FIG. 220 carries out a full-text search, and after obtaining the response result 110 and comparing the intent data 112 to generate the definite intent grammar data 114, and finally the analysis result 104 sent by the analysis result output module 116 to generate a return answer), and will select from the candidate list Select an answer more consistent with the first voice input 501 from all reported answers as the first reported answer 511 (for example, select an exact match record, etc.). Since the first reported answer 511 is the answer obtained by the internal analysis of the natural language understanding system 520, it must be converted into a voice output before being output to the user, so that the user can make a judgment. Then the speech synthesis module 526 will query the speech synthesis database 530 according to the first reported answer 511, and this speech synthesis database 530 is for example recorded with text and its corresponding voice information, which can make the speech synthesis module 526 find out the corresponding A first voice 513 that reports an answer 511 to synthesize a first voice response 507 . Afterwards, the voice synthesis module 526 can output the synthesized first voice response 507 to the user through a voice output interface (not shown) (wherein the voice output interface is, for example, a loudspeaker, a loudspeaker, or an earphone). It should be noted that when the speech synthesis module 526 queries the speech synthesis database 530 according to the first reported answer 511 , it may first need to convert the format of the first reported answer 511 , and then make a call through the interface specified by the speech synthesis database 530 . Whether format conversion is required when calling the speech synthesis database 530 is related to the definition of the speech synthesis database 530 itself, and this part is a technology well known to those skilled in the art, so it will not be described in detail here. the

Next enumerate examples to illustrate, if what the user inputs is the first voice input 501 of "I want to see Romance of the Three Kingdoms", the voice recognition module 522 will receive the first voice input 501 that is transmitted from the voice sampling module 510 and analyze it In the first request information 503, for example, the first keyword 509 containing "Romance of the Three Kingdoms" is taken out. The natural language processing module 524 can then analyze the first keyword 509 "Romance of the Three Kingdoms" (for example, perform a full-text search on the structured database 220 through the retrieval system 200 of FIG. After the comparison, generate the determined intent grammar data 114, and finally the analysis result 104 sent by the analysis result output module 116), and then generate the return answer containing three intention options of "Romance of the Three Kingdoms", and integrate it into a candidate list (assuming Each intent option has only one return answer, which are classified into three options of "reading", "watching TV series", and "watching movie" respectively), and then select one of the three return answers from the candidate list to be popular Field 316 has the highest value (eg pick record 10 of FIG. 3B ) as the first reported answer 511 . In one embodiment, the method corresponding to the highest value of the popularity field 316 can be directly executed (for example, directly playing "betrayal" sung by Xiao Jingteng to the user as mentioned earlier), and the present invention is not limited thereto. the

In addition, the natural language processing module 524 can also judge the previous first report by analyzing the subsequent received second speech input 501' (because it is fed into the speech sampling module 510 in the same way as the previous speech input 501). Is the answer 511 correct. Because the second voice input 501' is the user's response to the first voice response 507 previously provided to the user, it contains information about whether the user thinks the previous first voice response 507 is correct or not. If after analyzing the second voice input 501', it indicates that the user thinks that the first reported answer 511 is incorrect, the natural language processing module 524 can select other reported answers in the above-mentioned candidate list as the second reported answer 511', for example, from the candidate list After removing the first reported answer 511, and re-selecting a second reported answer 511' from the remaining reported answers, and then using the speech synthesis module 526 to find out the second voice 513' corresponding to the second reported answer 511', the ultimatum The voice synthesis module 526 synthesizes the second voice 513' into a second voice response 507' and plays it to the user. the

Continuing the example of the previous user inputting "I want to watch Romance of the Three Kingdoms", if the user wants to watch TV dramas of Romance of the Three Kingdoms, the option of record 10 in Figure 3B previously output to the user (because it is watching the movie "Romance of the Three Kingdoms") will be It is not what the user wants, so the user may input "I want to watch the Romance of the Three Kingdoms TV series" (the user clearly indicates that what he wants to watch is a TV series), or "I don't want to watch the Romance of the Three Kingdoms movie" (the user only denies the current option)...etc. The second voice input 501'. Then the second voice input 501' will analyze and obtain its second request information 503' (or second keyword 509'), and it will be found that the second keyword 509' in the second request information 503' will include " TV series" (the user has a clear instruction) or "no movie" (the user only negates the current option), so it will be judged that the first returned answer 511 does not meet the needs of the user. Therefore, at this time, another report answer can be selected from the candidate list as the second report answer 511' and the corresponding second voice response 507' is output, for example, the second voice response of "I will play the Romance of the Three Kingdoms TV series for you now" is output. Voice response 507' (if the user clearly indicates that he wants to watch the Romance of the Three Kingdoms TV series), or output the second voice response 507' of "which option you want" (if the user only negates the current option), combined with other options in the candidate list options for the user to select (for example, "select the return answer with the second highest numerical value in the popularity field 316 as the second return answer 511'). Moreover, in another embodiment, if the second voice input 501' input by the user includes " "Select" message, for example, when three options are displayed for the user to choose from: "Watch Romance of the Three Kingdoms Books", "Watch Romance of the Three Kingdoms TV Series", and "Watch Romance of the Three Kingdoms Movies", the user may input the second voice of "I want to watch a movie" When inputting 501', after analyzing the second request information 503' of the second voice input 501' and finding the user's intention (for example, finding that the user selects "watch a movie" from the second keyword 509'), the second voice input 501' will analyze and obtain its second request information 503', output the second voice response 507' of "I will play Romance of the Three Kingdoms movie for you now" (if the user wants to watch Romance of the Three Kingdoms movie) and then directly play the movie to the user. Of course, if what the user inputs is "I want the third option" (assuming that what the user chooses at this time is to read books), the application program corresponding to the third selection will be executed, that is, the output "what you want is Read the second voice response 507' of Romance of the Three Kingdoms book ", combined with the action of displaying the e-book of Romance of the Three Kingdoms to the user. 

In this embodiment, the speech recognition module 522 , the natural language processing module 524 and the speech synthesis module 526 in the aforementioned natural language understanding system 520 can be configured in the same machine as the speech sampling module 510 . In other embodiments, the speech recognition module 522 , the natural language processing module 524 and the speech synthesis module 526 may also be distributed in different machines (such as computer systems, servers or similar devices/systems). For example, in the natural language understanding system 520' shown in Figure 5C, the speech synthesis module 526 and the speech sampling module 510 can be configured in the same machine 502, while the speech recognition module 522 and the natural language processing module 524 can be configured in another machine. In addition, under the framework of FIG. 5C, the natural language processing module 524 will send the first reported answer 511/second reported answer 511' to the speech synthesis module 526, which then uses the first reported answer 511/second reported answer 511' Send to the speech synthesis database to find the corresponding first speech 513/second speech 513', as the basis for generating the first speech response 507/second speech response 507'. the

FIG. 6 is a flowchart of a method for modifying the first voice response 507 according to an embodiment of the present invention. In the method for modifying the first voice response 507 in this embodiment, when the user thinks that the currently played first voice response 507 does not conform to the first request information 503 that he previously input, he will input the second voice input 501 again. 'and fed into the voice sampling module 510, and then analyzed by the natural language understanding system 520 and learned that the first voice response 507 previously played to the user does not meet the user's intention, the natural language understanding system 520 can output the second voice again Response 507 ′, so as to modify the original first voice response 507 . For convenience of description, only the natural language dialogue system 500 in FIG. 5A is taken as an example here, but the method for modifying the first voice response 507 in this embodiment is also applicable to the natural language dialogue system 500' in FIG. 5C. the

Please refer to FIG. 5A and FIG. 6 at the same time. In step S602, the voice sampling module 510 receives the first voice input 501 (also fed into the voice sampling module 510). Wherein, the first voice input 501 is, for example, a voice from a user, and the first voice input 501 may also have the user's first request information 503 . Specifically, the first voice input 501 from the user may be an inquiry sentence, a command sentence, or other request information, such as "I want to watch Romance of the Three Kingdoms", "I want to listen to the music of Wangqingshui" or "what's the temperature today" etc. the

In step S604 , the natural language understanding system 520 analyzes at least one first keyword 509 included in the first voice input 501 to obtain a candidate list, wherein the candidate list has one or more reported answers. For example, when the user's first voice input 501 is "I want to watch Romance of the Three Kingdoms", the first keyword 509 obtained by the natural language understanding system 520 after analysis is, for example, ""Romance of the Three Kingdoms", "watch"" . For another example, when the user's first voice input 501 is "I want to listen to the song of Wangqingshui", the first keyword 509 obtained by the natural language understanding system 520 after analysis is, for example, ""Wangqingshui", "listen". ,"Song"". the

Then, the natural language understanding system 520 can query from the structured database 220 according to the above-mentioned first keyword 509, and obtain at least one search result (for example, the analysis result 104 in FIG. 1 ), as a return in the candidate list Answer. As for the method of selecting the first return answer 511 from multiple return answers, it can be described in FIG. 1A , and will not be repeated here. Since the first keyword 509 may contain different fields of knowledge (such as movies, books, music, or games, etc.), and the same field of knowledge can be further divided into multiple categories (such as different categories of the same movie or book title) authors, different singers of the same song name, different versions of the same game name, etc.), so for the first keyword 509, the natural language understanding system 520 can query one or more records related to The search results of the first keyword 509 (such as the analysis result 104), wherein each search result may include guidance data related to the first keyword 509 (such as "Hsiao Jingteng", "betrayal" as the keywords 108 When the structured database 220 in FIGS. 3A and 3B performs a full-text search, for example, two groups of matching results of records 6 and 7 in FIG. 3A will be obtained, which respectively include the guide data of "singerguid" and "songnameguid", which are stored in The data of the guide field 310) and other data. Wherein, other data are, for example, in the search results, other keywords etc. except those related to the first keyword 709 (for example, take "the days we walked together" as the keyword and do the full text in the structured database 220 of Fig. 3A When record 1 obtained from the retrieval is a matching result, "Andy Lau" and "Hong Kong and Taiwan, Cantonese, popular" are other data). Therefore, from another point of view, when the first voice input 501 input by the user has multiple first keywords 509, it means that the user's first request information 503 is relatively clear, so that the natural language understanding system 520 can better query Search results close to the first request information 503 . the

For example, when the first keyword 509 is "Romance of the Three Kingdoms" (for example, when the user enters the voice input of "I want to watch Romance of the Three Kingdoms"), the natural language understanding system 520 may generate three possible intent grammar data 106 after analysis ( As shown in Figure 1):

"<readbook>,<bookname>=Romance of the Three Kingdoms"; 

"<watchTV>,<TVname>=Romance of the Three Kingdoms"; and

"<watchfilm>,<filmname>=Romance of the Three Kingdoms". the

Therefore, the search results inquired are about "..."Romance of the Three Kingdoms"..."Book"" (the intention data is <readbook>), "..."Romance of the Three Kingdoms"..."TV Drama"" (the intention The data is <watchTV>), "..."Romance of the Three Kingdoms"..."movie"" (intent data is <watchfilm>) records (such as records 8, 9, and 10 in Figure 3B), where "TV drama" and "Books" and "Movies" respectively list the corresponding user intents). For another example, when the first keyword 509 is ""Wangqingshui", "music"" (for example, the user enters the voice input of "I want to listen to the music of Wangqingshui"), the natural language understanding system 520 may generate the following after analysis Possible intent syntax data:

"<playmusic>,<songname>=Forget Love Water"; 

The search results inquired are, for example, records about "..."Wangqingshui"..."Andy Lau"" (such as record 11 in Figure 3B), "..."Wangqingshui"..."Li Yijun"" (for example, record 12 in FIG. 3B ), wherein "Andy Lau" and "Li Yijun" are the intention data corresponding to the user. In other words, each search result may include the first keyword 509 and intent data related to the first keyword 509, and the natural language understanding system 520 will convert the data included in the search result according to the search result queried Generate a report answer, and record the report answer in the candidate list for use in subsequent steps. the

In step S606 , the natural language understanding system 520 selects at least one first reported answer 511 from the candidate list, and outputs a corresponding first voice response 507 according to the first reported answer 511 . In this embodiment, the natural language understanding system 520 can arrange the reported answers in the candidate list according to the priority, and select the reported answer from the candidate list according to the priority, so as to output the first voice response 507 . the

For example, when the first keyword 509 is "Romance of the Three Kingdoms", it is assumed that the natural language understanding system 520 has inquired many records about "..."Romance of the Three Kingdoms"..."book"" (that is, by querying The number of data received is prioritized), followed by the records of "..."Romance of the Three Kingdoms"..."Music"", and the number of records about "..."Romance of the Three Kingdoms"..."TV Drama"" At least, then the natural language understanding system 520 will use "the books of the Romance of the Three Kingdoms" as the first return answer (the most preferred return answer), and "the music of the Romance of the Three Kingdoms" as the second return answer (the second most preferred return answer) Answer), "The TV series of the Romance of the Three Kingdoms" as the third return answer (the third preferred return answer). Of course, if there is more than one record of the first return answer related to "Books of the Romance of the Three Kingdoms", the first return answer 511 can also be selected according to the priority order (for example, the number of clicks). The relevant details have been mentioned above. , which will not be described here. the

Next, in step S608, the voice sampling module 510 receives the second voice input 501', and the natural language understanding system 520 analyzes the second voice input 501', and determines whether the previously selected first reported answer 511 is correct. Here, the voice sampling module 510 will analyze the second voice input 501' to analyze the second keyword 509' included in the second voice input 501', wherein the second keyword 509' is, for example, further provided by the user. keywords (e.g. time, intent, domain of knowledge...etc.). And, when the second keyword 509' in the second voice input 501' does not match the relevant intent data in the first reported answer 511, the natural language understanding system 520 will determine that the previously selected first reported answer 511 is incorrect. As for judging whether the second request information 503' of the second voice input 501' contains "correct" or "negative" first voice response 507, it has been mentioned above, and will not be repeated here. the

Furthermore, the second speech input 501' parsed by the natural language understanding system 520 may or may not include an explicit second keyword 509'. For example, the voice sampling module 510 receives, for example, "I do not refer to the books of the Romance of the Three Kingdoms" (case A), "I do not refer to the books of the Romance of the Three Kingdoms, but I refer to the TV dramas of the Romance of the Three Kingdoms" (case A) from the user. B), "I mean the TV drama Romance of the Three Kingdoms" (case C) and so on. The second keyword 509' in the above case A is for example ""not", "Romance of the Three Kingdoms", "book"", and the keyword 509 in case B is for example "" not", "Romance of the Three Kingdoms", "book" , "Yes", "Romance of the Three Kingdoms", "TV drama"", and the second keyword 509' in the situation C is, for example, "" is ", "Romance of the Three Kingdoms", "TV drama"". For the convenience of description, the above-mentioned cases A, B and C are only cited as examples, but this embodiment is not limited thereto. the

Next, the natural language understanding system 520 will judge whether the relevant intention data in the first reported answer 511 is correct or not according to the second keyword 509' included in the second voice input 501'. That is to say, if the first return answer 511 is determined to be "Books of the Romance of the Three Kingdoms", and the above-mentioned second keyword 509' is ""Romance of the Three Kingdoms", "TV series"", the natural language understanding system 520 will determine the first return The relevant intent data in answer 511 (that is, the user wants to watch Romance of the Three Kingdoms "Book") does not match the second keyword 509' from the user's second voice input 501' (that is, the user wants to watch Romance of the Three Kingdoms "TV series"), so as to determine the One returns the answer 511 incorrect. Similarly, if the judgment return answer is "Books of the Romance of the Three Kingdoms", and the above-mentioned second keyword 509' is ""No", "Romance of the Three Kingdoms", "Book"", the natural language understanding system 520 will also judge that the first One returns the answer 511 incorrect. the

When the natural language understanding system 520 analyzes the second voice input 501 and judges that the previously output first voice response 501 is correct, then as shown in step S610, the natural language understanding system 520 will make a response corresponding to the second voice input 501′. response. For example, assuming that the second voice input 501' from the user is "Yes, it is a book about Romance of the Three Kingdoms", the natural language understanding system 520 may output the second voice response of "Helping you open the book about Romance of the Three Kingdoms" 507'. Alternatively, the natural language understanding system 520 can directly load the book content of Romance of the Three Kingdoms through a processing unit (not shown) while playing the second voice response 507'. the

However, when the natural language understanding system 520 analyzes the second voice input 501' and judges that the previously output first voice response 507 (that is, the reported answer 511) is incorrect, then as shown in step S612, the natural language understanding system 520 will Select another answer other than the first reported answer 511 from the candidate list, and output a second voice response 507' according to the selected result. Here, if the second voice input 501' provided by the user does not have a clear second keyword 509' (such as the second voice input 501' in the above case A), the natural language understanding system 520 can start from Select the second preferred return answer from the candidate list. Alternatively, if the second voice input 501' provided by the user has a clear second keyword 509' (such as the second voice input 501' in the above cases B and C), the natural language understanding system 520 can directly The second keyword 509' of the guide selects the corresponding return answer from the candidate list. the

On the other hand, if the second voice input 501' provided by the user has a clear second keyword 509' (such as the second voice input of the above-mentioned cases B and C), but the natural language understanding system 520 searches the candidate list If there is no returned answer matching the second keyword 509, the natural language understanding system 520 will output a third voice response, such as "I can't find this book" or "I don't know". the

In order for those skilled in the art to further understand the method for modifying voice response and the natural language dialogue system of this embodiment, another embodiment is given below for detailed description. the

First, assume that the first voice input 501 received by the voice sampling module 510 is "I want to watch Romance of the Three Kingdoms" (step S602), and then, the natural language understanding system 520 can analyze the first voice input 501 as ""watch", "Romance of the Three Kingdoms"" A keyword 509, and obtain a candidate list with a plurality of first reported answers, wherein each reported answer has related keywords and other data (other data can be stored in the content field 306 of Fig. 3A/3B, or each A part of the value field 312 of the record 302) (step S604), as shown in Table 1 (assuming that there is only one piece of data about the books/TV dramas/music/movies about Romance of the Three Kingdoms in the search results). the

Table I

Next, the natural language understanding system 520 will select the desired answer from the candidate list. Assuming that the natural language understanding system 520 sequentially selects the return answer a in the candidate list as the first return answer 511, the natural language understanding system 520, for example, outputs "whether to play the book of Romance of the Three Kingdoms", that is, the first voice response 507 ( Step S606). the

At this time, if the second voice input 501' received by the voice sampling module 510 is "Yes" (step S608), the natural language understanding system 520 will determine that the above-mentioned reported answer a is correct, and the natural language understanding system 520 will Output another voice response 507 "please wait" (that is, the second voice response 507'), and load the book content of Romance of the Three Kingdoms through a processing unit (not shown) (step S610). the

However, if the second voice input 501' received by the voice sampling module 510 is "I do not refer to the books of Romance of the Three Kingdoms" (step S608), the natural language understanding system 520 will judge that the above-mentioned return answer a is incorrect, and naturally The language understanding system 520 will then select another reported answer from the reported answers b-e in the candidate list as the second reported answer 511 ′, which is, for example, “whether to play the TV series Romance of the Three Kingdoms” for the reported answer b. If the user continues to answer "not a TV series", the natural language understanding system 520 will choose to report one of the answers c-e. In addition, if the reported answers a~e in the candidate list have been reported to the user by the natural language understanding system 520, and none of the reported answers a~e matches the user's voice input 501, the natural language understanding system 520 outputs "check No data" voice response 507 (step S612). the

In another embodiment, in the above-mentioned step S608, if the voice sampling module 510 receives the user's second voice input 501' as "I mean the comics of the Romance of the Three Kingdoms", here, since there is no comics in the candidate list Return the answer, so the natural language understanding system 520 will directly output the second voice response 507' of "no data found". the

Based on the above, the natural language understanding system 520 can output a corresponding first voice response 507 according to the first voice input 501 from the user. Wherein, when the first voice response 507 output by the natural language understanding system 520 does not conform to the request information 503 of the user's first voice input 501, the natural language understanding system 520 can modify the originally output first voice response 507, and according to the user's The subsequent second voice input 501 ′ further outputs a second voice response 507 ′ that is more in line with the user's first request information 503 . In this way, if the user is still dissatisfied with the answer provided by the natural language understanding system 520, the natural language understanding system 520 can automatically make corrections and report a new voice response to the user, so as to enhance the interaction between the user and the natural language dialogue system 500. Ease of conversation. the

It is worth mentioning that, in step S606 and step S612 in FIG. 6 , the natural language understanding system 520 can also sort the reported answers in the candidate list according to different methods of evaluating priority order, and then select the answer from the candidate list according to this priority order. Select the reported answer from among, and then output the voice response corresponding to the reported answer. the

For example, the natural language understanding system 520 can prioritize the first reported answers 511 in the candidate list according to the usage habits of the people, and the answers that are more frequently used by the people are prioritized. For example, taking the first keyword 509 as "Romance of the Three Kingdoms" as an example, assume that the returned answers found by the natural language understanding system 520 are TV dramas of Romance of the Three Kingdoms, books of Romance of the Three Kingdoms, and music of Romance of the Three Kingdoms. Among them, when people mention "Romance of the Three Kingdoms", they usually refer to the books of "Romance of the Three Kingdoms", less people refer to the TV series of "Romance of the Three Kingdoms", and even fewer people refer to the music of "Romance of the Three Kingdoms" (for example, use Figure 3C When the value stored in the heat field 316 in represents the matching situation of all users, the value of the heat field 316 will be the highest in the "book" record of "Romance of the Three Kingdoms", then the natural language understanding system 520 will sort according to the priority order about Return answers for "Books", "TV Dramas", and "Music". That is to say, the natural language understanding system 520 will preferentially select “Books of Romance of the Three Kingdoms” as the first reported answer 511 , and output the first voice response 507 according to the first reported answer 511 . the

In addition, the natural language understanding system 520 can also determine the priority order of the reported answers according to user habits. Specifically, the natural language understanding system 520 can record the speech input (including the first speech input 501, the second speech input 501', or any speech input by the user) once received from the user in the characteristic database ( For example, as shown in FIG. 7A/7B), wherein the characteristic database can be stored in a storage device such as a hard disk. The characteristic database can record the first keyword 509 obtained when the natural language understanding system 520 parses the user's voice input 501 , and the response records generated by the natural language understanding system 520 and other data about user preferences and habits. The storage and retrieval of user preference/habit data will be further described later with reference to FIGS. 7A/7B/8. In addition, in one embodiment, when the value stored in the popularity field 316 in FIG. 3C is related to the user's habit (eg, matching times), the user's usage habit or priority can be judged by the value of the popularity field 316 . Therefore, when the natural language understanding system 520 selects an answer to report, it can report answers in priority order according to information such as user habits recorded in the characteristic database 730 , so as to output a voice response 507 that is more in line with the user's voice input 501 . For example, in FIG. 3B , the values stored in the heat field 316 of records 8/9/10 are 2/5/8, respectively, which can respectively represent the "books", "television dramas" and "movies" of "Romance of the Three Kingdoms". " are 2/5/8 respectively, so the return answer corresponding to "Movies of the Romance of the Three Kingdoms" will be given priority. the

On the other hand, the natural language understanding system 520 can also choose to report an answer according to user habits. For example, suppose that when the user talks to the natural language understanding system 520, he often mentions "I want to read the books of the Romance of the Three Kingdoms", but seldom mentions "I want to watch the TV series of the Romance of the Three Kingdoms", and even less mentions "I want to watch the Romance of the Three Kingdoms". To watch the music of Romance of the Three Kingdoms" (for example, there are 20 records about "Books of Romance of the Three Kingdoms" in the user dialogue database (such as shown in the favorite field 318 of record 8 in Figure 3B), 8 records about "TV dramas of Romance of the Three Kingdoms" record (such as shown in the favorite field 318 of record 9 in Figure 3B), and a record about "Music of the Romance of the Three Kingdoms"), the priority order of the returned answers in the candidate list will be "Books of the Romance of the Three Kingdoms" , "TV Dramas of the Romance of the Three Kingdoms" and "Music of the Romance of the Three Kingdoms". That is to say, when the first keyword 509 is "Romance of the Three Kingdoms", the natural language understanding system 520 will select "Books of the Romance of the Three Kingdoms" as the first return answer 511, and output the first answer 511 based on the first return answer 511. Voice response 507. the

It is worth mentioning that the natural language understanding system 520 can also determine the priority order of the returned answers according to user preferences. Specifically, the user dialogue database can also record the keywords expressed by the user, for example: "like", "idol", "dislike" or "hate" and so on. Therefore, the natural language understanding system 520 can sort the reported answers from the candidate list according to the number of times the keywords are recorded. For example, if there are more times of "like" in the reported answer, the reported answer will be selected first. Or, if there are more times of "dislike" in the reported answer, it will be selected later. the

For example, assume that when a user talks to the natural language understanding system 520, he often mentions "I hate watching TV dramas of Romance of the Three Kingdoms", and seldom mentions "I hate listening to music of Romance of the Three Kingdoms", and even less mentions " I hate listening to the books of Romance of the Three Kingdoms" (for example, there are 20 records about "I hate watching TV dramas of Romance of the Three Kingdoms" in the user dialogue database (for example, it can be recorded through the disgust field 320 of record 9 in Figure 3B), 8 records about " I hate listening to the music of the Romance of the Three Kingdoms" record, and 1 record about "I hate reading the Romance of the Three Kingdoms" (for example, make a record through the disgust field 320 of record 8 in Figure 3B)), then the priority order of the return answer in the candidate list The order is "Books of the Romance of the Three Kingdoms", "TV Dramas of the Romance of the Three Kingdoms" and "Music of the Romance of the Three Kingdoms". That is to say, when the first keyword 509 is "Romance of the Three Kingdoms", the natural language understanding system 520 will select the book "Romance of the Three Kingdoms" as the first return answer 511, and output the first answer 511 based on the first return answer 511. Voice response 507. In an embodiment, a "dislike field 320" may be added in addition to the popularity field 316 in FIG. 3B to record the user's "dislike degree". In another embodiment, when the user's "dislike" information for a certain record is analyzed, one (or other numerical value) can be directly subtracted from the popularity field 316 (or favorite field 318) of the corresponding record, so that the Record user preference when not adding fields. Various implementations of recording user preferences can be applied in the embodiments of the present invention, and the present invention is not limited thereto. Other embodiments regarding the recording and application of user habit data, and the use habits and preferences of users/people to provide answers and return answers will be explained in more detail in FIG. 7A/7B/8 later. the

On the other hand, the natural language understanding system 520 can also be based on the fact that before the user provides the return answer earlier than the natural language dialogue system 500 (for example, before the first voice input 501 is played, the user does not yet know what kind of return the natural language dialogue system 500 will provide. answer for its choice) input voice input to determine the priority of at least one reported answer. That is to say, assuming that there is a voice input (for example, the fourth voice input) received by the voice sampling module 510 earlier than when the first voice input 501 is played, the natural language understanding system 520 can also analyze the fourth voice input the fourth keyword, and preferentially select a fourth reported answer that matches the fourth keyword in the candidate list, and output a fourth voice response based on the fourth reported answer. the

For example, assume that the natural language understanding system 520 first receives the first voice input 501 of "I want to watch a TV series", and not long after (for example, after a few seconds), it is assumed that the natural language understanding system 520 receives "Help me" again. Play the Romance of the Three Kingdoms." Input 501 in the fourth voice. At this time, the natural language understanding system 520 can recognize the first keyword 509 of "TV series" in the first voice input 501, and then recognize "Romance of the Three Kingdoms" in the fourth keyword. Therefore, the natural language understanding system 520 selects the reported answers about "Romance of the Three Kingdoms" and "TV drama" from the candidate list, and outputs a fourth voice response to the user based on the fourth reported answer. the

Based on the above, the natural language understanding system 520 can output a voice that is more in line with the voice input request information based on the voice input from the user and taking into account information such as everyone's usage habits, user preferences, user habits, or the user's previous and subsequent conversations. Reply to the user. Among them, the natural language understanding system 520 can prioritize the reported answers in the candidate list according to different sorting methods, such as people's usage habits, user preferences, user habits, or user conversations before and after. In this way, if the voice input from the user is relatively unclear, the natural language understanding system 520 can determine what the user's voice input 501 means by taking into account the usage habits of everyone, user preferences, user habits, or the previous and subsequent dialogues spoken by the user. intent (for example, attributes of keywords 509 in the first speech input 501, domain of knowledge, etc.). In other words, the natural language understanding system 520 will give priority to the reported answer if the reported answer is close to the intention expressed by the user or referred to by the public. In this way, the voice response output by the natural language dialogue system 500 can be more in line with the user's request information. the

To sum up, in the method for modifying voice response and the natural language dialogue system of this embodiment, the natural language dialogue system can output the corresponding first voice response 507 according to the first voice input 501 from the user. Wherein, when the first voice response 507 output by the natural language dialogue system does not conform to the first request information 503 or the first keyword 509 of the user's first voice input 501, the natural language dialogue system can correct the originally output first voice Response 507, and further select a second voice response 507' that better meets the user's needs according to the second voice input 501' provided by the user subsequently. In addition, the natural language dialogue system can also select a more appropriate return answer based on the usage habits of the people, user preferences, user habits, or the preceding and following conversations of the user, and then output the corresponding voice response to the user. In this way, if the user is not satisfied with the answer provided by the natural language dialogue system, the natural language dialogue system can automatically correct it according to the request information spoken by the user each time, and report a new voice response to the user, so as to enhance the user experience. Convenience when having a conversation with a natural language dialogue system. the

Next, the structure and components of the natural language understanding system 100 and the structured database 220 are used to illustrate the example of providing answers and reporting answers based on the dialogue scene and context with the user, the user's usage habits, everyone's usage habits, and user preferences. . the

FIG. 7A is a block diagram of a natural language dialogue system according to an embodiment of the present invention. Referring to FIG. 7A , the natural language dialogue system 700 includes a speech sampling module 710 , a natural language understanding system 720 , a feature database 730 and a speech synthesis database 740 . In fact, the speech sampling module 710 in FIG. 7A is the same as the speech sampling module 510 in FIG. 5A , and the natural language understanding system 520 is also the same as the natural language understanding system 720 , so the functions they perform are the same. In addition, when the natural language understanding system 720 analyzes the request information 703, it can also obtain the user's intention by performing a full-text search on the digitized database 220 in FIG. Let me repeat. As for the characteristic database 730 is used to store the user preference data 715 sent by the natural language understanding system 720 or provide the user preference record 717 to the natural language understanding system 720, this part will be described in detail later. The speech synthesis database 740 is equivalent to the speech synthesis database 530 and is used to provide speech output to the user. In this embodiment, the voice sampling module 710 is used to receive the voice input 701 (that is, the first/second voice input 501/501' in FIG. 5A/B is the voice from the user), and the natural language understanding system 720 will analyze The request information 703 in voice input (that is, the first/second request information 503/503' of FIG. 5A/B), and output the corresponding voice response 707 (that is, the first/second voice response 507/ 507'). Each component in the aforementioned natural language dialogue system 700 may be configured in the same machine, which is not limited in the present invention. the

The natural language understanding system 720 will receive the request information 703 after analyzing the speech input 701 transmitted from the speech sampling module 710, and the natural language understanding system 720 will generate one or more keywords 709 according to the speech input 701 A candidate list containing at least one reported answer, and then find out from the candidate list one that is more in line with the keyword 709 as the reported answer 711, and query the speech synthesis database 740 accordingly to find out the voice 713 corresponding to the reported answer 711, and finally Then output a voice response 707 according to the voice 713 . In addition, the natural language understanding system 720 of this embodiment can be implemented by a hardware circuit composed of one or several logic gates, or implemented by computer program codes, which is only an example and not limited thereto. . the

Fig. 7B is a block diagram of a natural language dialogue system 700' according to another embodiment of the present invention. The natural language understanding system 720' in FIG. 7B may include a speech recognition module 722 and a natural language processing module 724, and the speech sampling module 710 and the speech synthesis module 726 may be combined in a speech synthesis processing module 702. Wherein, the voice recognition module 722 receives the request information 703 for analyzing the voice input 701 from the voice sampling module 710 and converts it into one or more keywords 709 . The natural language processing module 724 then processes the keywords 709 to obtain at least one candidate list, and selects one that is more consistent with the voice input 701 from the candidate list as the return answer 711 . Since the reported answer 711 is the answer obtained by the internal analysis of the natural language understanding system 720, it must be converted into text or voice output before it can be output to the user, so the speech synthesis module 726 will query the speech synthesis database 740 according to the reported answer 711 , and the speech synthesis database 740 is, for example, recorded text and its corresponding speech information, which enables the speech synthesis module 726 to find the speech 713 corresponding to the reported answer 711 to synthesize the speech response 707 . Afterwards, the speech synthesis module 726 can output the synthesized speech through a speech output interface (not shown), wherein the speech output interface is, for example, a loudspeaker, a loudspeaker, or an earphone, etc.), so as to output the speech to the user. It should be noted that in FIG. 7A , the natural language understanding system 720 incorporates a speech synthesis module 726 (such as the architecture of FIG. 5B , but the speech synthesis module 726 is not shown in FIG. 7A ), and the speech synthesis module will utilize The report answer 711 queries the speech synthesis database 740 to obtain the speech 713 as a basis for synthesizing the speech response 707 . the

In this embodiment, the speech recognition module 722, the natural language processing module 724 and the speech synthesis module 726 in the aforementioned natural language understanding system 720 can be respectively equivalent to the speech recognition module 522, the natural language processing module 524 and the speech synthesis module of FIG. 5B Module 526 and provides the same functionality. In addition, the speech recognition module 722, the natural language processing module 724 and the speech synthesis module 726 can be configured in the same machine as the speech sampling module 710. In other embodiments, the speech recognition module 722 , the natural language processing module 724 and the speech synthesis module 726 may also be distributed in different machines (such as computer systems, servers or similar devices/systems). For example, in the natural language understanding system 720' shown in FIG. 7B, the speech synthesis module 726 and the speech sampling module 710 can be configured in the same machine 702, while the speech recognition module 722 and the natural language processing module 724 can be configured in another machine. It should be noted that, in the framework of FIG. 7B, since the speech synthesis module 726 and the speech sampling module 710 are configured in a machine 702, the natural speech understanding system 720 needs to transmit the return answer 711 to the machine 702, and synthesized by speech The module 726 sends the reported answer 711 to the speech synthesis database 740 to find the corresponding speech 713 as a basis for generating the speech response 707 . In addition, when the speech synthesis module 726 calls the speech synthesis database 740 according to the reported answer 711, it may first need to convert the format of the returned answer 711, and then make a call through the interface specified by the speech synthesis database 740, because this part belongs to the technology in the art The technology is well known to the personnel, so it will not be described in detail here. the

The natural language dialogue method will be described below in conjunction with the above natural language dialogue system 700 in conjunction with FIG. 7A . FIG. 8 is a flowchart of a natural language dialogue method according to an embodiment of the present invention. For convenience of description, only the natural language dialog system 800 in FIG. 7A is taken as an example here, but the natural language dialog method of this embodiment can also be applied to the natural language dialog system 700' in FIG. 7B. Compared with Figure 5/6, Figure 5/6 deals with automatically correcting the output information based on the user's voice input, but Figure 7A/7B/8 deals with recording user preference data based on the characteristic database 730 715, and select one from the candidate list as the answer 711 accordingly, and play its corresponding voice to the user. In fact, the implementations of Fig. 5/6 and Fig. 7A/7B/8 can be selected or coexisted, and the invention is not limited thereto. the

Please refer to FIG. 7A and FIG. 8 at the same time. In step S810 , the voice sampling module 710 receives the voice input 701 . Wherein, the voice input 701 is, for example, voice from a user, and the voice input 701 may also have request information 703 of the user. Specifically, the voice input 701 from the user can be an inquiry sentence, a command sentence, or other request information, such as the examples mentioned above "I want to watch the Romance of the Three Kingdoms", "I want to listen to the music of Wangqingshui" or "today's temperature How many times" and so on. It should be noted that steps S802-S806 are the process for the natural language dialogue system 700 to store user preference data 715 for the user's previous voice input, and subsequent steps S810-S840 are based on these user preference data previously stored in the characteristic database 730 715 to operate. The details of steps S802-S806 will be described later, and the operation content of steps S820-S840 will be described below. the

In step S820, the natural language understanding system 720 analyzes at least one keyword 709 included in the first voice input 701, and then obtains a candidate list, wherein the candidate list has one or more reported answers. In detail, the natural language understanding system 720 analyzes the speech input 701 to obtain one or more keywords 709 of the speech input 701 . For example, when the user's voice input 701 is "I want to watch Romance of the Three Kingdoms", the keyword 709 obtained by the natural language understanding system 720 after analysis is, for example, ""Romance of the Three Kingdoms", "watch"" (as mentioned above description, and then analyze whether the user wants to watch a book, a TV series, or a movie). For another example, when the user's voice input 701 is "I want to listen to the song of Wangqingshui", the keyword 709 obtained after analysis by the natural language understanding system 720 is, for example, ""Wangqingshui", "listen", "song" "(As mentioned earlier, it can be further analyzed that what the user wants to listen to is the version sung by Andy Lau or Li Yijun). Then, the natural language understanding system 720 can perform a full-text search from the structured database according to the above-mentioned keywords 709, and obtain at least one search result (which can be at least one record in FIG. 3A/3B), and use it as a candidate Return answers in a list. Since a keyword 709 may belong to different fields of knowledge (such as movies, books, music or games, etc.), and the same field of knowledge can be further divided into multiple categories (such as different authors of the same movie or book title , different singers of the same song title, different versions of the same game title, etc.), so for a keyword 709, the natural language understanding system 720 can obtain a One or more search results related to the keyword 709, which include other information except the keyword 709 and the keyword 709 (the content of other information is shown in Table 1). Therefore, from another point of view, when the first voice input 701 input by the user has multiple keywords 709, it means that the user's request information 703 is relatively clear, so that the natural language understanding system 720 can analyze the request information 703 more clearly. Close search results (because if the natural language understanding system 720 can find a complete match result, it should be the option that the user wants). the

For example, when the keyword 709 is "Romance of the Three Kingdoms", the search results analyzed by the natural language understanding system 720 are, for example, about "..."Romance of the Three Kingdoms"..."TV drama"", "..." Records of "Romance of the Three Kingdoms"..."Books" (where "TV series" and "Books" are the user intentions indicated by the response results). For another example, when the keyword 709 is ""Wangqingshui", "music"", the user's intent analyzed by the natural language understanding system 720 may be "..."Wangqingshui"..."music"... Records of "Andy Lau", "..."Wangqingshui"..."Music"..."Li Yijun"", where "Andy Lau" and "Li Yijun" are the search results used to indicate the user's intention. In other words, after the natural language understanding system 720 performs a full-text search on the structured database 220, each search result may include keywords 709 and other data related to the keywords 709 (as shown in Table 1), and the natural language understanding The system 720 converts the analyzed search results into a candidate list including at least one reported answer for subsequent steps. the

In step S830, the natural language understanding system 720 uses the user preference record 717 sent from the characteristic database 730 (for example, the result collected based on the user preference data 715 stored therein, which will be described later), to select from the candidate Select a reported answer 711 from the list, and output a voice response 707 according to the reported answer 711 . In this embodiment, the natural language understanding system 720 can select the reported answer 711 from the candidate list according to a priority order (how the priority order includes will be described in detail below). In step S840, according to the reported answer 711, a voice response 707 is output (step S840). the

For example, in one embodiment, the number of search results can be prioritized. For example, when the keyword 709 is "Romance of the Three Kingdoms", it is assumed that the natural language understanding system 720 finds in the structured database 220 about ". .."Romance of the Three Kingdoms"..."Books"" has the largest number of records, followed by records of "..."Romance of the Three Kingdoms"..."Music"", and about "..."Romance of the Three Kingdoms".. "TV series"" has the least number of records, then the natural language understanding system 720 will use the records related to "Romance of the Three Kingdoms" as the first priority return answer (for example, organize all books about "Romance of the Three Kingdoms" into a candidate list , and can be sorted according to the value of the popularity field 316), records related to "Romance of the Three Kingdoms" are used as the second priority return answer, and records related to "Romance of the Three Kingdoms" are used as the third priority return answer. It should be noted that, in addition to the number of search results, the priority may also be based on user preferences, user habits, or public usage habits, and related descriptions will be described in detail later. the

In order for those skilled in the art to further understand the natural language dialogue method and the natural language dialogue system of this embodiment, another embodiment is given below for detailed description. the

First, assume that the first voice input 701 received by the voice sampling module 710 is "I want to watch the Romance of the Three Kingdoms" (step S810), and then, the natural language understanding system 720 can parse out the key words of ""watch", "Romance of the Three Kingdoms"" word 709, and obtain a candidate list with a plurality of reported answers, wherein each reported answer has related keywords (step S820) and other information, as shown in Table 1 above. the

Next, the natural language understanding system 720 selects a reward answer from the candidate list. Assuming that the natural language understanding system 720 selects the return answer a in the candidate list (please refer to Table 1) as the first return answer 711, then the natural language understanding system 720, for example, outputs "whether to play the book of Romance of the Three Kingdoms" as a voice response 707 (steps S830-S840). the

As mentioned above, the natural language understanding system 720 can also sort the reported answers in the candidate list according to different evaluation priority methods, and output the voice response 707 corresponding to the reported answer 711 accordingly. For example, the natural language understanding system 720 can determine the user's preference based on multiple conversation records with the user (such as using the user's positive/negative words mentioned above), that is, the user preference record 717 can be used to determine the return answer 711 priority. However, before explaining how users use positive/negative terms, let us explain how the user preference data 715 stores preferences/dislikes or habits of users/everyone. the

Now follow steps S802-806 regarding the storage method of the user preference data 715. In one embodiment, before receiving the voice input 701 in step S810, multiple voice inputs, that is, previous historical conversation records may be received in step S802, and user preferences may be retrieved based on these multiple previous voice inputs 701 The data 715 (step S804 ) is then stored in the property database 730 . In fact, the user preference data 715 can also be stored in the structured database 220 (or in a manner of merging the characteristic database 730 into the structured database 220). For example, in one embodiment, the popularity field 316 in FIG. 3B can be directly used to record the user's preferences. As for the recording method of the popularity field 316, it has been mentioned above (for example, when a certain record 302 is matched, its popularity field will be added 1), which will not be repeated here. Of course, it is also possible to create another field in the structured database 220 to store the user preference data 715, such as using keywords (such as "Romance of the Three Kingdoms") as the basis, combined with user preferences (such as when the user mentions positive terms such as "like" and For negative terms such as "disgusting", you can add one to the values in the favorite field 318 and the disgusting field 320 in FIG. Therefore, when the natural language understanding system 720 queries the user preference record 717 in the structured database 200, it can directly query the values of the preference field 318 and the dislike field 320 (you can query the number of positive terms and negative terms), and then make a judgment based on the The user's preferences (that is, the statistics of positive and negative terms are sent to the natural language understanding system 720 as the user preference record 717). the

The following will describe the situation where the user preference information 715 is stored in the property database 730 (that is, the property database 730 is not incorporated into the structured database 220). In one embodiment, the user preference information 715 can be stored using a keyword corresponding to the user's "favorite" for the keyword. For example, the storage of the user preference information 715 can directly use the preference field 852 and The dislike field 862 is used to record the personal likes and dislikes of a certain keyword, and the like field 854 and the dislike field 864 are used to record the likes and dislikes of the group of keywords. For example, in FIG. 8B , the values of the favorite field 852 and the dislike field 862 corresponding to the keyword ""Romance of the Three Kingdoms", "book"" stored in the record 832 are 20 and 1 respectively, and the keyword "" stored in the record 834 The values of the preference field 852 and the dislike field 862 corresponding to "Romance of the Three Kingdoms" and "TV drama" are respectively 8 and 20, and the preference fields corresponding to the keywords "Romance of the Three Kingdoms" and "Music"" stored in record 836 The numerical values of 852 and dislike field 862 are respectively 1 and 8, which all represent the user's personal preference and dislike data for relevant keywords (for example, the higher the numerical value of the liking field 852, the more he likes, and the higher the numerical value of the dislike field 862, the more disgust). In addition, the values of the like field 854 and the dislike field 864 corresponding to the record 832 are 5 and 3 respectively, the values of the like field 854 and the dislike field 864 corresponding to the record 834 are 80 and 20 respectively, and the like field 854 corresponding to the record 836 The numerical value of field 864 and dislike is respectively 2 and 10, which represent the likes and dislikes data (abbreviated as "favorite instruction") of the public for related keywords, so the like field 852 and dislike can be increased according to user's preferences The value of field 862. Therefore, if the user inputs the voice of "I want to watch the TV series of the Romance of the Three Kingdoms", the natural language understanding system 720 can combine the "keywords" ""Romance of the Three Kingdoms", "TV series"" with the "favorite indication" that increases the value of the preference field The user preference data 715 is sent to the characteristic database 730, so the characteristic database 730 can add 1 to the value of the preference field 852 of the record 834 (because the user wants to watch ""Romance of the Three Kingdoms" and "TV series"", it means that the degree of preference has increased ). According to the above-mentioned method of recording user preference data, when the user enters relevant keywords again in the future, for example, when the user inputs "I want to watch Romance of the Three Kingdoms", the natural language understanding system 720 can use the keyword "Romance of the Three Kingdoms" in the picture The characteristic database 730 of 8B finds three records 832/834/836 related to "Romance of the Three Kingdoms", and the characteristic database 730 can return the values of the favorite field 852 and the dislike field 862 as the user preference record 717 to the natural language understanding system 720, so the natural language understanding system 720 can use the user preference record 717 as the basis for judging the user's personal preference. Of course, the characteristic database 730 can also return the values of the like field 854 and the dislike field 864 as the user preference record 717 to the natural language understanding system 720, but at this time the user preference record 717 will be used as a basis for judging everyone's preferences. The user preferences record 717 represents the preferences of the user or the public and is not limited. the

In another embodiment, the values of the like field 852 and the dislike field 862 can also be used as a basis for judging user/people habits. For example, after receiving the user preference record 717, the natural language understanding system 720 can first judge the numerical difference between the like field 852/854 and the dislike field 862/864, if the difference between the two values exceeds a certain critical value, it means that the user It is customary to use a specific method for dialogue. For example, when the value of the favorite field 852 is more than 10 times greater than the value of the dislike field 862, it means that the user particularly likes to use "positive terms" for dialogue (this is a record of "user habits") mode), so the natural language understanding system 720 can only use the preference field 852 to select the reported answer in this situation. When the natural language understanding system 720 uses the value of the like field 854/dislike field 864 stored in the characteristic database 730, it means that what is judged is the preference records of all users in the characteristic database 730, and the judgment result can be used as a basis for everyone's usage habits. reference data. It should be noted that the user preference record 717 sent back from the characteristic database 730 to the natural language understanding system 720 may include both the user's personal preference record (such as the value of the preference field 852/dislike field 862) and the public preference record (such as the preference field 854/value of dislike field 864), the present invention does not limit this. the

As for the storage of the user preference data 715 obtained based on the voice input this time, when the candidate list is generated in step S820 (whether it is a complete match or a partial match), the natural language dialog system 700 stores the data in the user voice input this time. The obtained user preference data 715 . For example, in step S820, whenever a keyword can generate a matching result in the structured database 220, it can be determined that the user has a preference for the matching result, so the "keyword" and "favorite indication" can be sent to Go to the characteristic database 730, and after finding the corresponding record therein, change the corresponding preference field 852/854 or dislike field 862/864 value of the corresponding record (for example, when the user inputs "I want to see the books of the Romance of the Three Kingdoms", the The value of the preference field 852/854 of record 832 of FIG. 8B is incremented by one). In yet another embodiment, the natural language dialogue system 700 may also store the user preference data 715 after the user selects a report answer in step S830. In addition, if the corresponding keyword is not found in the characteristic database 730 , a new record can be created to store the user preference data 715 . For example, when the user inputs the voice of "I listen to Andy Lau's Wangqingshui" and generates keywords ""Andy Lau", "Wangqingshui"", if the corresponding record is not found in the characteristic database 730 when storing, it will be stored in the characteristic database 730 Create a new record 838, and add one to the value of the corresponding preference field 852/854. The storage timing and storage method of the above-mentioned user preference data 715 are for illustration purposes only, and those skilled in the art may change the embodiments shown in the present invention according to actual applications, but all equivalent modifications that do not depart from the spirit of the present invention should still be included in the claims of the present invention. the

In addition, although the format of storing records 832-838 in the characteristic database 730 shown in FIG. 8B is different from the record format of the structured database 220 (such as those shown in FIGS. The storage format is not limited. Moreover, although the above-mentioned embodiment only describes the storage and use of the like field 852/854 and the dislike field 862/864, in another embodiment, another field 872/874 can be set up in the characteristic database 730 to store the user/dislike field 872/874 respectively. Other habits of the public, such as the number of downloads, citations, recommendations, comments, or referrals of the data corresponding to this record. In another embodiment, the times or data of these downloads, citations, recommendations, comments, or referrals can also be stored in the favorite field 852/854 and the dislike field 862/864, for example, each time a user provides When a good comment or referral is given to others for reference, one can be added to the value of the favorite field 852/854. If the user provides a bad comment to a certain record, one can be added to the value of the dislike field 862/864. There is no limit to the number of records and the method of recording the values of the fields. It should be noted that those skilled in the art should know that since the like field 852 and dislike field 862 in FIG. Stored in the user's mobile communication device, and the like field 854, dislike field 864... and other information related to all users are stored in the server, thus saving the storage space of the server and keeping the privacy of the user's personal preferences sex. the

The actual usage status of the user will be further described below using FIG. 7A and FIG. 8B . Based on the dialogue content of multiple voice inputs 701, it is assumed that when the user talks to the natural language understanding system 720, he often mentions "I hate watching TV dramas of the Romance of the Three Kingdoms" and seldom mentions "I hate listening to the music of the Romance of the Three Kingdoms", And mention less "I hate listening to the books of Romance of the Three Kingdoms" (for example, there are 20 records about "TV dramas of Romance of the Three Kingdoms I hate watching" in the characteristic database 730 (that is, in Fig. 8B, "Romance of the Three Kingdoms" plus " The number of negative terms of TV series" is 20), 8 records about "I hate listening to the music of Romance of the Three Kingdoms" (that is, in Figure 8B, the number of negative terms of "Romance of the Three Kingdoms" plus "music" is 8), And 1 about " I hate to listen to the books of Romance of the Three Kingdoms ") (that is, in Fig. 8B, the quantity of the negative term of " Romance of the Three Kingdoms " plus " book " is 1), because the user preference returned from characteristic database 730 Record 717 will contain the quantity of these three negative terms (that is, 20, 8, 1), then the natural language understanding system 720 will arrange the priority order of the return answer 711 in the candidate list as "Books of the Romance of the Three Kingdoms", "The Music of the Romance of the Three Kingdoms", and "The TV Drama of the Romance of the Three Kingdoms". That is to say, when the keyword 709 is "Romance of the Three Kingdoms", the natural language understanding system 720 will select the book "Romance of the Three Kingdoms" as the return answer 711, and output a voice response 707 based on the return answer 711. It should be noted that although the above is to use the statistical value of the negative words used by the user alone to set the priority order, in another embodiment, the statistical value of the positive words used by the user can still be used alone to set the priority order ( For example, as mentioned earlier, the value of the like field 852 is higher than a certain threshold value of the dislike field 862). the

It is worth mentioning that the natural language understanding system 720 can also determine the priority order of the returned answers based on the amount of positive words and negative words used by the user. Specifically, the characteristic database 730 can also record the keywords expressed by the user, such as: "like", "idol" (the above are positive terms), "dislike" or "hate" (the above are negative terms), etc. . Therefore, the natural language understanding system 720, in addition to comparing the difference between the number of times the user uses "like" and "dislike", can also directly evaluate the return answer based on the number of times of positive/negative terms corresponding to the above keywords from the candidate list. Ranking (i.e. comparing positive terms or negative terms which has more citations). For example, if there are more times of "like" in the reported answer (that is, more positive words are cited, or the value of the favorite field 852 is larger), then this reported answer will be selected preferentially. Or, if there are more times related to "dislike" in the reported answer (that is, the number of references to negative terms is larger, or the numerical value of the disgust field 862 is relatively large), then it will be selected later, so the natural language understanding system 720 can All the reported answers are sorted into a candidate list according to the above priority order. Since some users may prefer to use positive terms (for example, the numerical value of the favorite field 852 is particularly large), while other users are accustomed to using negative terms (such as the numerical value of the dislike field 862 is particularly large), so in the above-mentioned embodiment, because the user preference record 717 will reflect the usage habits of individual users, so options more in line with user habits can be provided for them to choose. the

In addition, the natural language understanding system 720 can also sort the priority order of the reported answers 711 in the candidate list according to the usage habits of the people, and the answers that are more frequently used by the people are prioritized (for example, use the heat field 316 in FIG. 3C to record ). For example, when the keyword 709 is "Romance of the Three Kingdoms", it is assumed that the returned answer found by the natural language understanding system 720 is, for example, TV dramas of Romance of the Three Kingdoms, books of Romance of the Three Kingdoms, and music of Romance of the Three Kingdoms. Among them, when people mention "Romance of the Three Kingdoms", they usually refer to TV dramas of "Romance of the Three Kingdoms", fewer people refer to movies of "Romance of the Three Kingdoms", and even fewer people refer to books of "Romance of the Three Kingdoms", (such as Figure 8B , the values of the relevant records in the preference field 854 are 80, 40, and 5), then the natural language understanding system 720 will sort the returned answers 711 about "tv series", "movie" and "book" according to the order of priority. That is to say, the natural language understanding system 720 will preferentially select "TV series of Romance of the Three Kingdoms" as the return answer 711 , and output a voice response 707 based on the return answer 711 . As for the above-mentioned method of "arranging the answers frequently used by everyone first", you can use the popularity field 316 in FIG. 3C to record, and the recording method has been disclosed in the relevant paragraphs of FIG. 3C above, and will not be repeated here. the

In addition, the natural language understanding system 720 may also determine the priority order of the reported answers 711 according to the frequency of use of the user. Specifically, because the natural language understanding system 720 can record the voice input 701 received from the user in the characteristic database 730, the characteristic database 730 can record the keywords obtained when the natural language understanding system 720 parses the user's voice input 701 709 and all response information such as the return answer 711 generated by the natural language understanding system 720. Therefore, when the natural language understanding system 720 selects the return answer 711 in the future, according to the response information recorded in the characteristic database 730 (such as user preferences/dislikes/habits, or even everyone's preferences/dislikes/habits...), according to the priority Sort to find the reward answer 711 that is more in line with the user's intention (determined by the user's voice input), so as to provide a corresponding voice response. As for the above method of "determining the priority order of the reported answer 711 according to the user's habit", the popularity field 316 in FIG. 3C can also be used for recording, and the recording method has been disclosed in the relevant paragraph of FIG. 3C above, and will not be repeated here. the

Based on the above, the natural language understanding system 720 can store the above-mentioned user preference attributes (such as positive terms and negative terms), user habits, and public usage habits in the characteristic database 730 (step S806 ). That is to say, in step S802, step S804, and step S806, the user preference data 715 is obtained from the user's previous historical conversation records, and the collected user preference data 715 is added to the characteristic database 730. Habits and people's usage habits are stored in the characteristic database 730, so that the natural language understanding system 720 can use the rich information in the characteristic database 730 (such as user preference records 717) to provide users with more correct answers. the

Next, the details of step S830 will be further described. In step S830, after the voice input is received in step S810, and the keyword 709 of the voice input is analyzed in S820 to obtain a candidate list, then the natural language understanding system 720 uses user preferences such as user preferences, user habits, or everyone's usage habits, etc. Record 717, determine the priority order of at least one reported answer (step S880). As mentioned above, prioritization can be based on the number of records searched, positive/negative terms by the user or crowd, etc. Next, select a report answer 711 from the candidate list according to the priority order (step S890 ). As for the selection of the report answer, the one with the highest matching degree or the one with the highest priority order can be selected as described above. Afterwards, according to the reported answer 711, a voice response 707 is output (step S840). the

On the other hand, the natural language understanding system 720 can also determine the priority order of at least one reported answer according to the voice input 701 input earlier by the user. That is to say, assuming that another voice input 701 (such as the aforementioned fourth voice input) is received by the voice sampling module 710 earlier than the voice response 707 is played, the natural language understanding system 720 can also analyze the The keyword (that is, the fourth keyword 709) in this voice input 701 (that is, the fourth voice input), and in the candidate list, preferentially select the return answer that matches the keyword as the return answer 711, and Output a voice response 707 according to the reported answer 711 . the

For example, suppose the natural language understanding system 720 first receives the voice input 701 of "I want to watch a TV series", and after a few seconds, suppose the natural language understanding system 720 receives the voice input 701 of "Help me play Romance of the Three Kingdoms" Voice input 701. At this time, the natural language understanding system 720 can recognize the keyword (first keyword) of "TV series" in the first voice input 701, and recognize the keyword of "Romance of the Three Kingdoms" (the fourth keyword) later. ), therefore, the natural language understanding system 720 will select from the candidate list, the intention data is the return answer about "Romance of the Three Kingdoms" and "TV drama", and report the answer 711 so as to output the voice response 707 to the user. the

Based on the above, the natural language understanding system 720 can output the request information 703 that is more in line with the voice input 701 based on the voice input from the user, and taking into account information such as everyone's usage habits, user preferences, user habits, or the conversations that the user said before and after. The voice response 707 is given to the user. Among them, the natural language understanding system 720 can prioritize the reported answers in the candidate list according to different sorting methods, such as people's usage habits, user preferences, user habits, or user conversations before and after. In this way, if the voice input 701 from the user is relatively unclear, the natural language understanding system 720 can determine what the user's voice input 701 means by taking into account the usage habits of everyone, user preferences, user habits, or the conversations before and after the user speaks. Refers to the intent (such as the attribute of the keyword 709 in the speech input, the domain of knowledge, etc.). In other words, if the reported answer 711 is close to the intention expressed by the user or referred to by the public, the natural language understanding system 720 will give priority to the reported answer 711 . In this way, the voice response 707 output by the natural language dialogue system 700 can be more in line with the user's request information 703 . the

It should be noted that although the characteristic database 730 and the structured database 220 are described above as different databases, these two databases can be combined together, and those skilled in the art can choose according to actual applications. the

To sum up, the present invention provides a natural language dialogue method and its system. The natural language dialogue system can output a corresponding voice response according to the voice input from the user. The natural language dialog system of the present invention can also preferentially select a more appropriate return answer based on the usage habits of the people, user preferences, user habits, or the previous and subsequent dialogues spoken by the user, so as to output a voice response to the user, thereby Improve the convenience of users when they have a dialogue with the natural language dialogue system. the

Then, the structure and components of the natural language understanding system 100 and the structured database 220 are applied to the number of returned answers obtained by analyzing the request information input by the user's voice, and the decision is made to directly operate based on the data type or require the user to provide further information. instructions, and then when there is only one answer left in the report, it can also be explained directly based on the example of the data type. The advantage of providing users with this choice is that the system does not need to screen the reported answers for the user, but directly provides the user with a candidate list containing the reported answers, and allows the user to decide the software or What kind of service is provided to achieve the purpose of providing a user-friendly interface. the

FIG. 9 is a system diagram of a mobile terminal device according to an embodiment of the present invention. 9, in this embodiment, the mobile terminal device 900 includes a voice receiving unit 910, a data processing unit 920, a display unit 930 and a storage unit 940. The data processing unit 920 is coupled to the voice receiving unit 910 , the display unit 930 and the storage unit 940 . The voice receiving unit 910 is used for receiving the first input voice SP1 and the second input voice SP2 and sending them to the data processing unit 920 . The aforementioned first voice input SP1 and second voice input SP2 may be voice inputs 501 and 701 . The display unit 930 is controlled by the data processing unit 920 to display the first/second candidate list 908/908'. The storage unit 940 is used for storing a plurality of data, which may include the data of the aforementioned structured database 220 or the characteristic database 730 , which will not be repeated here. In addition, the storage unit 940 can be any type of memory in the server or computer system, such as dynamic random access memory (DRAM), static random access memory (SRAM), flash memory (Flash memory), read-only memory (ROM)...etc. , the present invention is not limited to this, and those skilled in the art can select according to actual needs. the

In this embodiment, the function of the data processing unit 920 is similar to that of the natural language understanding system 100 in FIG. Process to generate the first keyword 904 corresponding to the first input speech SP1, and according to the first keyword 904 corresponding to the first input speech SP1 from the data of the storage unit 940 (such as the search engine 240 according to the keyword 108 to the structured database 220 Find the first reported answer 906 (for example, the first reported answer 511/711) in the full-text search). When the number of first reported answers 906 found is 1, the data processing unit 920 can directly perform corresponding operations according to the type of data corresponding to the first reported answers 906; The unit 920 can organize the first reported answer 906 into a first candidate list 908, and then control the display unit 940 to display the first candidate list 908 to the user. In the situation where the first candidate list 908 is displayed for the user to make further selections, the data processing unit 920 will receive the second input voice SP2, and perform voice recognition on it to generate the second request information 902', and then process the second request information 902' performs natural language processing to generate a second keyword 904' corresponding to the second input speech SP2, and selects a corresponding part from the first candidate list 908 according to the second keyword 904' corresponding to the second input speech SP2. Wherein, the first keyword 904 and the second keyword 904' may be composed of multiple keywords. The above method of analyzing the second voice input SP2 to generate the second request information 902' and the second keyword 904' can use the method of analyzing the second voice input shown in Figs. 5A and 7A, so it will not be repeated here. the

Similarly, when the quantity of the second reported answer 906 is 1, the data processing unit 920 will perform corresponding operations according to the type of the second reported answer 906; when the quantity of the second reported answer 906' is greater than 1, the data processing unit 920 A second candidate list 908' will be organized according to the second reported answer 906' and displayed by controlling the display unit 940. Then, select the corresponding part according to the next voice input by the user, and then perform the corresponding operation according to the number of subsequent reported answers. the

Further, the data processing unit 920 will compare the multiple records 302 of the structured database 220 (such as the numerical data of each sub-field 308 in the title field 304) with the first keyword 904 corresponding to the first input speech SP1 (As previously described for Figures 1, 3A, 3B, 3C). When a certain record 302 of the structured database 220 is at least partially matched with the first keyword 904 of the first input speech SP1, the record 302 is regarded as the matching result produced by the first input speech SP1 (for example, FIG. 3A/3B produces matching results). Wherein, if the type of data is a music file, then record 302 may include song name, artist, album name, publication time, play order, ... etc.; if the type of data is video file, then record 302 may include movie name, publication time , staff (including performers), ... etc.; if the type of data is a web page file, the record 302 may include the website name, type of web page, corresponding user account, ... etc.; if the type of data is a picture file, record 302 may include picture name, picture information, . The above record 302 is an example for illustration, and the record 302 may be determined according to actual application, and the embodiment of the present invention is not limited thereto. the

Next, the data processing unit 920 can determine whether the second keyword 904' corresponding to the second input speech SP2 contains a sequence of words indicating the sequence (such as "I want the third option" or "I choose the third"). When the second keyword 904' corresponding to the second input speech SP2 contains a sequential vocabulary indicating the sequence, the data processing unit 920 selects the data at the corresponding position from the first candidate list 908 according to the sequential vocabulary. When the second keyword 904' corresponding to the second input voice SP2 does not contain the sequential vocabulary indicating the order, it means that the user may directly select a first reported answer 906 in the first candidate list 908, and the data processing unit 920 will convert the first The record 302 corresponding to each first reported answer 906 in the candidate list 908 is compared with the second keyword 904' to determine the degree of correspondence between the first reported answer 906 and the second input speech SP2, and then determine the first key word according to the degree of correspondence. Whether there is a certain first reported answer 906 in the candidate list 908 corresponds to the second input speech SP2. In an embodiment of the present invention, the data processing unit 920 can determine the corresponding degree of the first reported answer 906 to the second keyword 904' (such as the degree of complete match or partial match) to determine the first candidate list 906 Whether there is a first reported answer 906 corresponding to the second input voice SP2, so as to simplify the selection process. Wherein, the data processing unit 920 may select the one with the highest corresponding degree among the data as corresponding to the second input speech SP2. the

For example, if the first input speech SP1 is "what's the weather today", after performing speech recognition and natural language processing, the first keyword 904 corresponding to the first input speech SP1 will include "today" and "weather", so The data processing unit 920 reads the data corresponding to today's weather, and displays the weather data as the first candidate list 908 through the display unit 930 . Then, if the second input voice SP2 is "I want to see the third data" or "I choose the third data", after voice recognition and natural language processing, the second keyword 904' corresponding to the second input voice SP2 Can comprise " the 3rd pen ", here " the 3rd pen " can be interpreted as the sequential vocabulary of indication order, so the data processing unit 920 can read the 3rd pen data in the first candidate list 908 (that is the first candidate list 908 in the third first return answer 906), and then display the corresponding weather information through the display unit 930. Or, if the second input voice SP2 is "I want to see the weather in Beijing" or "I choose the weather in Beijing", after voice recognition and natural language processing, the second keyword 904' corresponding to the second input voice SP2 will be Include "Beijing" and "weather", so the data processing unit 920 will read the data corresponding to Beijing in the first candidate list 908 . When the number of the first reported answer 906 corresponding to this selection is 1, the corresponding weather information can be displayed directly through the display unit 930; when the number of the selected first reported answer 906 is greater than 1, further second candidates are displayed The list 908' (including at least one second reported answer 906') is for the user to further select. the

If the first input speech SP1 is "I want to call Lao Zhang", after performing speech recognition and natural language processing, the first keyword 904 corresponding to the first input speech SP1 will include "telephone" and "Zhang", so The data processing unit 920 will read the contact person data corresponding to the surname "Zhang" (the detailed data corresponding to the record 302 can be obtained by performing a full-text search on the structured database 220), and display these contact person data through the display unit 930 A first candidate list 908 of data (ie, a first reported answer 906). Next, if the second input speech SP2 is "the third old Zhang" or "I choose the third", after performing speech recognition and natural language processing, the second keyword 904' corresponding to the second input speech SP2 will include "The 3rd", here "the 3rd" will be interpreted as a sequential vocabulary indicating the order, so the data processing unit 920 will read the 3rd data in the first candidate list 908 (that is, the third first Return answer 906), and dial according to the selected data. Or, if the second input speech SP2 is "I choose the one starting with 139", after speech recognition and natural language processing, the second keyword 904' corresponding to the second input speech SP2 will include "139" and "beginning", Here "139" will not be interpreted as a sequential vocabulary indicating the order, so the data processing unit 920 will read the contact data whose phone number starts with 139 in the first candidate list 908; if the second input voice SP2 is "I Want Beijing's Lao Zhang", after performing speech recognition and natural language processing, the second keyword 904' corresponding to the second input voice SP2 will include "Beijing" and "Zhang", and the data processing unit 920 will read the first candidate The contact person data whose address is Beijing in the list 908. When the selected first return answer 906 quantity is 1, then dial according to the selected data; The answers are reported 906' and sorted into a second candidate list 908' for the user to choose. the

If the first input voice SP1 is "I am looking for a restaurant", after performing voice recognition and natural language processing, the first keyword 904 of the first input voice SP1 will include "restaurant", and the data processing unit 920 will read all corresponding For the restaurant's first return answer 906, since such an instruction is not very clear, the first candidate list 908 (comprising the first return answer 906 corresponding to all restaurant data) will be displayed to the user through the display unit 930, and the user will wait for further instructions. Next, if the user inputs "the third restaurant" or "I choose the third" through the second input voice SP2, after performing voice recognition and natural language processing, the second keyword 904' corresponding to the second input voice SP2 Will include "the 3rd", where "the 3rd" will be interpreted as a sequential vocabulary indicating the order, so the data processing unit 920 will read the 3rd data in the first candidate list 908, and according to the selected data to display. Or, if the second input speech SP2 is "I choose the nearest", after performing speech recognition and natural language processing, the second keyword 904' corresponding to the second input speech SP2 will include "the nearest", so the data processing unit 920 will read the restaurant data whose address is closest to the user in the first candidate list 908; if the second input voice SP2 is "I want a restaurant in Beijing", after voice recognition and natural language processing, the second input voice SP2 corresponds to The second keyword 904 ′ includes “Beijing” and “restaurant”, so the data processing unit 920 will read the restaurant data whose address is Beijing in the first candidate list 908 . When the number of the selected first reported answer 906 is 1, it will be displayed according to the selected data; when the selected first reported answer 906 is greater than 1, then the selected first reported answer 906 will be used as the second report Answers 906', and sorted into a second candidate list 908' for the user to choose. the

According to the above, the data processing unit 920 can perform corresponding operations according to the data type of the selected first reported answer 906 (or the second reported answer 906'). For example, when the type of data corresponding to the selected first reported answer 906 is a music file, the data processing unit 920 performs music playback according to the selected data; The unit performs video playback according to the selected data; when the type of the selected data is a web page file, the data processing unit 920 displays according to the selected data; when the type of the selected data is a picture file, the data processing unit 920 The picture is displayed according to the selected data; when the type of the selected data is a card file, the data processing unit 920 dials according to the selected data. the

FIG. 10 is a system diagram of an information system according to an embodiment of the present invention. Please refer to FIG. 9 and FIG. 10. In this embodiment, the information system 1000 includes a mobile terminal device 1010 and a server 1020, wherein the server 1020 can be a cloud server, a local area network server, or other similar devices, but this The embodiments of the invention are not limited thereto. The mobile terminal device 1010 includes a voice receiving unit 1011 , a data processing unit 1013 and a display unit 1015 . The data processing unit 1013 is coupled to the voice receiving unit 1011 , the display unit 1015 and the server 1020 . The mobile terminal device 1010 may be a mobile communication device such as a cell phone (Cell phone), a personal digital assistant (Personal Digital Assistant, PDA), a smart phone (Smart phone), and the present invention is not limited thereto. The function of the voice receiving unit 1011 is similar to that of the voice receiving unit 910 , and the function of the display unit 1015 is similar to that of the display unit 930 . The server 1020 is used for storing a plurality of data and has a voice recognition function. the

In this embodiment, the data processing unit 1013 performs speech recognition on the first input speech SP1 through the server 1020 to generate the first request information 902, and then performs natural language processing on the first request information 902 to generate the corresponding first input speech The first keyword 904 of SP1, and the server 1020 will perform a full-text search on the structured database 220 according to the first keyword 904 to find out the first reported answer 906 and send it to the data processing unit 1013 . When the number of the first reported answer 906 is 1, the data processing unit 1013 will perform corresponding operations according to the data type corresponding to the first reported answer 906; when the number of the first reported answer 906 is greater than 1, the data processing unit 1013 will At this time, the selected first reported answer 906 is sorted into a first candidate list 908 and then the display unit 1015 is controlled to display to the user, and further instructions from the user are awaited. After the user inputs the instruction again, the data processing unit 1013 will perform voice recognition on the second input voice PS2 through the server 1020 to generate the second request information 902', and then analyze and process the second request information 902' with natural language To generate the second keyword 904' corresponding to the second input speech SP2, and the server 1020 selects the corresponding first reported answer 906 from the first candidate list 908 according to the second keyword 904' corresponding to the second input speech SP2 as The second returns the answer 906 ′ and sends it to the data processing unit 1013 . Similarly, when the number of the corresponding second reported answer 906 is 1, the data processing unit 920 will perform corresponding operations according to the type of data corresponding to the second reported answer 906; when the number of the second reported answer 906 is greater than At 1, the data processing unit 1013 will organize the second reported answer 906 selected at this time into a second candidate list 908 ′, and then control the display unit 1015 to display it for the user to make further choices. Then, the server 1020 will select the corresponding part according to the subsequent input voice, and the data processing unit 1013 will perform corresponding operations according to the quantity of the selected data. This can be obtained by analogy with reference to the above description, and will not be repeated here. . the

It should be noted that, in one embodiment, if the number of the first reported answer 906 selected according to the first keyword 904 corresponding to the first input voice SP1 is 1, the operation corresponding to the data can be performed directly. In addition, in another embodiment, a prompt may be first output to the user to inform the user that the corresponding operation of the selected first report answer 906 will be executed. Furthermore, in yet another embodiment, when the number of the second reported answer 906 selected according to the second keyword 904' corresponding to the second input voice SP2 is 1, the operation corresponding to the data can be directly performed. Certainly, in another embodiment, a prompt may also be first output to the user to inform the user that the corresponding operation of the selected data will be executed, and the present invention is not limited thereto. the

Further, the server 1020 compares each record 302 of the structured database 220 with the first keyword 904 corresponding to the first input speech SP1. When each record 302 at least partially matches the first keyword 904 , the record 302 is regarded as data matched by the first input speech SP1 , and the record 302 is used as one of the first reported answers 906 . If the number of first reported answers 906 selected according to the first keyword 904 corresponding to the first input voice SP1 is greater than 1, the user may input instructions through the second input voice SP2. Since the instructions input by the user through the second input voice SP2 at this time may include order (to indicate the order of which item in the displayed information to be selected), directly select a certain one in the displayed information (for example, to directly indicate a certain item of information) content), or determine the user’s intention according to the instructions (such as selecting the nearest restaurant, the “nearest” restaurant will be displayed to the user), so the server 1020 will then determine the second keyword corresponding to the second input voice SP2 904' contains a sequential word indicating the sequence. When the second keyword 904' corresponding to the second input voice SP2 contains a sequential vocabulary indicating the order, the server 1020 selects the first reported answer 906 at the corresponding position from the first candidate list 908 according to the sequential vocabulary. When the second keyword 904' corresponding to the second input speech SP2 does not contain the sequence vocabulary indicating the sequence, the server 1020 returns each first answer 906 in the first candidate list 908 to the second keyword corresponding to the second input speech SP2. The keyword 904' is compared to determine the degree of correspondence between the first reported answer 906 and the second input speech SP2, and it can be determined based on these corresponding degrees whether the first reported answer 906 in the first candidate list 908 corresponds to the second input speech SP2 . In an embodiment of the present invention, the server 1020 may determine which first reported answers 906 in the first candidate list 908 correspond to the second input speech SP2 according to the degree of correspondence between the first reported answers 906 and the second keyword 904′, to simplify the selection process. Wherein, the server 1020 may select the one with the highest corresponding degree among the first reported answers 906 as the one corresponding to the second input voice SP2. the

FIG. 11 is a flowchart of a selection method based on speech recognition according to an embodiment of the present invention. Please refer to FIG. 11 , in this embodiment, the first input voice is received (step S1100), and voice recognition is performed on the first input voice SP1 to generate the first request information 902 (step S1110), and then the first request information 902 Perform analysis of natural language processing to generate a first keyword 904 corresponding to the first input speech (step S1120 ). Next, the corresponding first reported answer 906 is selected from multiple data according to the first keyword 904 (step S1130 ), and it is determined whether the number of the selected first reported answer 906 is 1 (step S1140 ). When the number of the selected first reported answer 906 is 1, that is, the judgment result of step S1140 is “Yes”, then the corresponding operation is performed according to the data type corresponding to the first reported answer 906 (step S1150 ). When the number of the selected first reported answer 906 is greater than 1, that is, the judgment result of step S1140 is "No", the first candidate list 908 is displayed according to the selected first reported answer 906 and the second input voice SP2 is received (step S1160) , and perform voice recognition on the second input voice to generate the second request information 902' (step S1170), then analyze and process the second request information 902' to generate the second keyword 904 corresponding to the second input voice '(step S1180). Next, select the corresponding part from the first reported answer 906 in the first candidate list 908 according to the second request information 902, and return to step S1140 to determine whether the number of the selected first reported answer 906 is 1 (step S1190). Wherein, the sequence of the above steps is for illustration, and the embodiment of the present invention is not limited thereto. Moreover, the details of the above steps can refer to the embodiment shown in FIG. 9 and FIG. 10 , which will not be repeated here. the

In summary, the voice recognition-based selection method and its mobile terminal device and information system in the embodiment of the present invention perform voice recognition and natural language processing on the first input voice and the second input voice to confirm the first input voice and the second input voice. The keywords corresponding to the second input voice, and then select the returned answer according to the keywords corresponding to the first input voice and the second input voice. Thereby, the convenience of the user's operation can be improved. the

Next, an operation example of combining the structure and components of the natural language understanding system 100 and the structured database 220 disclosed in the present invention with the auxiliary starting device will be described. the

FIG. 12 is a block diagram of a voice control system according to an embodiment of the present invention. Referring to FIG. 12 , the voice control system 1200 includes an auxiliary activation device 1210 , a mobile terminal device 1220 and a server 1230 . In this embodiment, the auxiliary activation device 1210 activates the voice system of the mobile terminal device 1220 through wireless transmission signals, so that the mobile terminal device 1220 communicates with the server 1230 according to the voice signal. the

In detail, the auxiliary starting device 1210 includes a first wireless transmission module 1212 and a trigger module 1214 , wherein the trigger module 1214 is coupled to the first wireless transmission module 1212 . The first wireless transmission module 1212 supports, for example, wireless compatibility certification (Wireless fidelity, Wi-Fi), Worldwide Interoperability for Microwave Access (WiMAX), Bluetooth (Bluetooth), ultra-wideband (ultra-wideband, UWB) or radio-frequency identification (Radio-frequency identification, RFID) and other communication protocol devices, which can send out wireless transmission signals to correspond with another wireless transmission module to establish a wireless connection. The trigger module 1214 is, for example, a button, a key, and the like. In this embodiment, when the user presses the trigger module 1214 to generate a trigger signal, the first wireless transmission module 1212 receives the trigger signal and activates. A wireless transmission module 1212 transmits the wireless transmission signal to the mobile terminal device 1220 . In one embodiment, the above-mentioned auxiliary starting device 1210 may be a bluetooth earphone. the

It should be noted that although some hands-free earphones/microphones are also designed to activate certain functions of the mobile terminal device 1220, in another embodiment of the present invention, the auxiliary activation device 1210 may be different from the aforementioned earphone/microphone. The above-mentioned earphone/microphone is connected to the mobile terminal device to replace the earphone/microphone on the mobile terminal device 1220 for listening/talking. The activation function is an additional design, but the auxiliary activation device 1210 of the present invention is "only" used When the voice system in the mobile terminal device 1220 is turned on, it does not have the function of listening/talking, so the internal circuit design can be simplified, and the cost is also low. In other words, compared to the above-mentioned hands-free earphone/microphone, the auxiliary activation device 1210 is another device, that is, the user may have both the hands-free earphone/microphone and the auxiliary activation device 1210 of the present invention. the

In addition, the shape of the above-mentioned auxiliary starting device 1210 can be articles that are readily available to the user, such as rings, watches, earrings, necklaces, glasses and other decorations, that is, various portable items, or installation components, such as The driving accessories arranged on the steering wheel are not limited to the above. That is to say, the auxiliary activation device 1210 is a "living" device, through the setting of the internal system, the user can easily touch the trigger module 1214 to activate the voice system. For example, when the auxiliary activation device 1210 is in the form of a ring, the user can easily move his finger to press the trigger module 1214 of the ring to be triggered. On the other hand, when the shape of the auxiliary activation device 1210 is a device configured on a vehicle accessory, the user can easily trigger the trigger module 1214 of the vehicle accessory device during driving. In addition, compared to the uncomfortable feeling of wearing earphones/microphones for listening/talking, using the auxiliary starting device 1210 of the present invention can turn on the voice system in the mobile terminal device 1220, and even turn on the sound amplification function (details will be described later) ), so that the user can still listen/talk directly through the mobile terminal device 1220 without wearing an earphone/microphone. In addition, for users, these "life-like" auxiliary activation devices 1210 are articles that they would wear or use originally, so there will be no problem of unaccustomed or uncomfortable feeling in use, that is, no need to spend Time to adapt. For example, when a user is cooking in the kitchen and needs to make a call to a mobile phone placed in the living room, assuming he wears the auxiliary starting device 1210 of the present invention having a shape of a ring, necklace or watch, he can lightly touch the ring, necklace, etc. or watch to turn on the voice system to ask friends for recipe details. Although some earphones/microphones with activation function can also achieve the above-mentioned purpose, but in the process of cooking every time, it is not necessary to make a phone call to ask friends for advice every time, so for users, wearing earphones/microphones at any time It can be said that it is quite inconvenient to use a microphone for cooking in order to control the mobile terminal device at any time. the

In other embodiments, the auxiliary starting device 1210 can also be configured with a wireless rechargeable battery 1216 for driving the first wireless transmission module 1212 . Furthermore, the wireless charging battery 1216 includes a battery unit 12162 and a wireless charging module 12164 , wherein the wireless charging module 12164 is coupled to the battery unit 12162 . Here, the wireless charging module 12164 can receive energy supplied by a wireless power supply device (not shown), and convert the energy into electricity to charge the battery unit 12162 . In this way, the first wireless transmission module 1212 of the auxiliary starting device 1210 can be conveniently charged by the wireless rechargeable battery 1216 . the

On the other hand, the mobile terminal device 1220 is, for example, a mobile phone (Cell phone), a personal digital assistant (Personal Digital Assistant, PDA) mobile phone, a smart phone (Smart phone), or a palmtop computer (Pocket PC) with communication software installed. ), tablet PC (Tablet PC) or notebook computer, etc. The mobile terminal device 1220 can be any portable (Portable) mobile device with a communication function, and its scope is not limited here. In addition, the mobile terminal device 1220 may use an Android operating system, a Microsoft operating system, an Android operating system, a Linux operating system, etc., and is not limited to the above. the

The mobile terminal device 1220 includes a second wireless transmission module 1222, the second wireless transmission module 1222 can be matched with the first wireless transmission module 1212 of the auxiliary starting device 1210, and adopts a corresponding wireless communication protocol (such as wireless compatible certification, global Interoperability with communication protocols such as microwave access, bluetooth, ultra-wideband communication protocol or radio frequency identification), so as to establish a wireless connection with the first wireless transmission module 1212 . It should be noted that the “first” wireless transmission module 1212 and the “second” wireless transmission module 1222 mentioned here are used to illustrate that the wireless transmission modules are configured in different devices, and are not intended to limit the present invention. the

In other embodiments, the mobile terminal device 1220 further includes a voice system 1221, and the voice system 1221 is coupled to the second wireless transmission module 1222. Therefore, after the user triggers the trigger module 1214 of the auxiliary starting device 1210, the voice system 1221 can be transmitted through the first wireless transmission module. The module 1212 and the second wireless transmission module 1222 wirelessly activate the voice system 1221 . In one embodiment, the speech system 1221 may include a speech sampling module 1224 , a speech synthesis module 1226 and a speech output interface 1227 . The voice sampling module 1224 is used for receiving voice signals from the user. The voice sampling module 1224 is, for example, a device for receiving audio such as a microphone (Microphone). The speech synthesis module 1226 can query a speech synthesis database, and the speech synthesis database is, for example, information recorded with text and its corresponding speech, so that the speech synthesis module 1226 can find out the speech corresponding to a specific text message, so as to convert the text message into speech synthesis. Afterwards, the speech synthesis module 1226 can output the synthesized speech through the speech output interface 1227, so as to play it to the user. The aforementioned voice output interface 1227 is, for example, a speaker or an earphone. the

In addition, the mobile terminal device 1220 can also be configured with a communication module 1228 . The communication module 1228 is, for example, a component capable of transmitting and receiving wireless signals, such as a radio frequency transceiver. Furthermore, the communication module 1228 enables the user to receive or make calls or use other services provided by the telecommunications operator through the mobile terminal device 1220 . In this embodiment, the communication module 1228 can receive the response information from the server 1230 through the Internet, and establish a call connection between the mobile terminal device 1220 and at least one electronic device according to the response information, wherein the electronic device For example, it is another mobile terminal device (not shown). the

The server 1230 is, for example, a web server or a cloud server, which has a speech understanding module 1232 . In this embodiment, the speech understanding module 1232 includes a speech recognition module 12322 and a speech processing module 12324 , wherein the speech processing module 12324 is coupled to the speech recognition module 12322 . Here, the voice recognition module 12322 receives the voice signal from the voice sampling module 1224 to convert the voice signal into multiple semantic segments (such as keywords or sentences, etc.). The voice processing module 12324 can analyze the meanings (such as intention, time, location, etc.) represented by these segmentation semantics according to these segmentation semantics, and then determine the meaning expressed in the above-mentioned voice signal. In addition, the speech processing module 12324 will also generate corresponding response information according to the parsed results. In this embodiment, the speech understanding module 1232 may be implemented by a hardware circuit composed of one or several logic gates, or may be implemented by a computer program code. It is worth mentioning that, in another embodiment, the voice understanding module 1232 can be configured in a mobile terminal device 1320 , such as the voice control system 1300 shown in FIG. 13 . The operation of the speech understanding module 1232 of the above-mentioned server 1230 can be as shown in the natural language understanding system 100 of Figure 1A and the natural language dialogue system 500/700/700' of Figures 5A/7A/7B. the

The method of voice control will be described below in conjunction with the above-mentioned voice control system 1200 . FIG. 14 is a flowchart of a voice control method according to an embodiment of the present invention. Please refer to FIG. 12 and FIG. 14 at the same time. In step S1402 , the auxiliary starting device 1210 sends a wireless transmission signal to the mobile terminal device 1220 . In detail, when the first wireless transmission module 1212 of the auxiliary starting device 1210 is triggered by receiving a trigger signal, the auxiliary starting device 1210 will send a wireless transmission signal to the mobile terminal device 1220 . Specifically, when the trigger module 1214 in the auxiliary starting device 1210 is pressed by the user, the trigger module 1214 will be triggered by the trigger signal at this time, so that the first wireless transmission module 1212 will send a wireless transmission signal to the mobile terminal device 1220. The second wireless transmission module 1222 enables the first wireless transmission module 1212 to connect with the second wireless transmission module 1222 through a wireless communication protocol. The above-mentioned auxiliary activation device 1210 is only used to activate the voice system in the mobile terminal device 1220, and does not have the function of listening/talking, so the internal circuit design can be simplified and the cost is low. In other words, compared to the hands-free earphone/microphone attached to the general mobile terminal device 1220, the auxiliary activation device 1210 is another device, that is, the user may have both the hands-free earphone/microphone and the auxiliary activation device 1210 of the present invention. the

It is worth mentioning that the shape of the above-mentioned auxiliary starting device 1210 can be an article that is readily available to the user, such as various portable items such as rings, watches, earrings, necklaces, glasses, etc., or an installation component, such as The driving accessories arranged on the steering wheel are not limited to the above. That is to say, the auxiliary starting device 1210 is a "lifelike" device, through the setting of the internal system, the user can easily touch the trigger module 1214 to activate the voice system 1221 . Therefore, the voice system 1221 in the mobile terminal device 1220 can be turned on by using the auxiliary starting device 1210 of the present invention, and even the sound amplification function can be turned on (will be described in detail later), so that the user can still use the audio system without wearing an earphone/microphone Listen/call directly through the mobile terminal device 1220 . In addition, for the user, these "living-oriented" auxiliary activation devices 1210 are items that would be worn or used originally, so there will be no problem of unaccustomed or uncomfortable feeling in use. the

In addition, both the first wireless transmission module 1212 and the second wireless transmission module 1222 can be in sleep mode or working mode. Wherein, the sleep mode refers to that the wireless transmission module is in a closed state, that is, the wireless transmission module does not receive/detect wireless transmission signals, and cannot be connected with other wireless transmission modules. The working mode means that the wireless transmission module is in an open state, that is, the wireless transmission module can continuously detect wireless transmission signals, or send wireless transmission signals at any time, and can be connected with other wireless transmission modules. Here, when the trigger module 1214 is triggered, if the first wireless transmission module 1212 is in sleep mode, the trigger module 1214 will wake up the first wireless transmission module 1212, make the first wireless transmission module 1212 enter the working mode, and make the second A wireless transmission module 1212 sends a wireless transmission signal to the second wireless transmission module 1222, so that the first wireless transmission module 1212 is connected with the second wireless transmission module 1222 of the mobile terminal device 1220 through a wireless communication protocol. the

On the other hand, in order to prevent the first wireless transmission module 1212 from continuing to consume too much power in the working mode, within a preset time (for example, 5 minutes) after the first wireless transmission module 1212 enters the working mode, if the trigger If 1214 is not triggered again, the first wireless transmission module 1212 will enter the sleep mode from the working mode, and stop connecting with the second wireless transmission module 1220 of the mobile terminal device 1220 . the

After that, in step S1404 , the second wireless transmission module 1222 of the mobile terminal device 1220 receives the wireless transmission signal to activate the voice system 1221 . Next, in step S1406, when the second wireless transmission module 1222 detects a wireless transmission signal, the mobile terminal device 1220 can start the voice system 1221, and the 1221 sampling module 1224 of the voice system can start to receive voice signals, such as "What is the temperature today?" Degree", "Call Lao Wang.", "Please check the phone number." and so on. the

In step S1408, the voice sampling module 1224 transmits the voice signal to the voice understanding module 1232 in the server 1230, so that the voice understanding module 1232 analyzes the voice signal and generates response information. Further, the speech recognition module 12322 in the speech understanding module 1232 will receive the speech signal from the speech sampling module 1224, and divide the speech signal into a plurality of segmentation semantics, and the speech processing module 12324 will perform the above segmentation semantics Speech understanding to generate response information in response to speech signals. the

In another embodiment of the present invention, the mobile terminal device 1220 can further receive the response information generated by the voice processing module 12324 , so as to output the content of the response information through the voice output interface 1227 or execute the operation issued by the response information. In step S1410, the speech synthesis module 1226 of the mobile terminal device 1220 receives the response information generated by the speech understanding module 1232, and performs speech synthesis according to the content (such as words or sentences) in the response information to generate a speech response. And, in step S1412, the voice output interface 1227 receives and outputs the voice response. the

For example, when the user presses the trigger module 1214 in the auxiliary activation device 1210, the first wireless transmission module 1212 will send a wireless transmission signal to the second wireless transmission module 1222, so that the mobile terminal device 1220 activates the voice of the voice system 1221 Sampling module 1224 . Here, assuming that the voice signal from the user is an inquiry sentence, such as "what's the temperature today", the voice sampling module 1224 will receive and send the voice signal to the voice understanding module 1232 in the server 1230 for analysis, And the speech understanding module 1232 can send the response information generated by the analysis back to the mobile terminal device 1220 . Assuming that the content of the response information generated by the speech understanding module 1232 is "30°C", the speech synthesis module 1226 will synthesize the message of "30°C" into a speech response, and the speech output interface 1227 can broadcast the speech response to user. the

In another embodiment, assuming that the voice signal from the user is a command sentence, such as "call Lao Wang.", then the voice understanding module 1232 can recognize that the command sentence is a request for "call Lao Wang." ". In addition, the speech understanding module 1232 will generate a new response message, such as "please confirm whether to call Lao Wang", and send the new response message to the mobile terminal device 1220 . Here, the voice synthesis module 1226 synthesizes the new response information into a voice response, and broadcasts it to the user through the voice output interface 1227 . Furthermore, when the user's answer is an affirmative answer of "yes", similarly, the voice sampling module 1224 can receive and transmit the voice signal to the server 1230 for the voice understanding module 1232 to analyze. After the speech understanding module 1232 completes the analysis, it will record a dial instruction message in the response message and send it to the mobile terminal device 1220 . At this time, the communication module 1228 will query the phone number of "Lao Wang" according to the contact information recorded in the phone database, so as to establish a call connection between the mobile terminal device 1220 and another electronic device, that is, dial to "Old King". the

In other embodiments, in addition to the above-mentioned voice control system 1200 , the voice control system 1300 or other similar systems can also be used to perform the above-mentioned operation method, which is not limited to the above-mentioned embodiments. the

To sum up, in the voice control system and method of this embodiment, the auxiliary activation device can wirelessly activate the voice function of the mobile terminal device. Moreover, the shape of this auxiliary starting device can be "life-like" articles that are readily available to the user, such as rings, watches, earrings, necklaces, glasses and other decorations, that is, various portable items, or installation components , for example, a driving accessory arranged on the steering wheel, not limited to the above. In this way, it will be more convenient to use the auxiliary activation device 1210 of the present invention to activate the voice system in the mobile terminal device 1220 than the discomfort of wearing a hands-free earphone/microphone at present. the

It should be noted that the above-mentioned server 1230 with the speech understanding module may be a network server or a cloud server, and the cloud server may involve the issue of privacy of the user. For example, the user needs to upload the complete address book to the cloud server in order to complete operations related to the address book such as making calls and sending text messages. Even if the cloud server adopts an encrypted connection, and the instant transmission is not saved, it is still difficult to eliminate the user's worries. Accordingly, another method of voice control and its corresponding voice interaction system are provided below. The mobile terminal device can perform voice interaction services with the cloud server without uploading the complete address book. In order to make the content of the present invention clearer, the following specific examples are given as examples in which the present invention can actually be implemented. the

Although the present invention has been disclosed above with embodiments, it is not intended to limit the present invention. Those skilled in the art can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection is based on the claims of the present invention. the

Claims (20)

1. system of selection based on speech recognition comprises:

Receive one first phonetic entry;

This first phonetic entry is carried out speech recognition to produce one first key word;

This first key word is carried out natural language processing to produce one first user view that should first phonetic entry;

Select at least one first repayment answer according to this first user view;

When the quantity of this first repayment answer of selecting is 1, carry out corresponding operation according to the type of selected this first repayment answer;

When the quantity of this first repayment answer of selecting greater than 1 the time, show that one comprises first candidate data of this first repayment answer;

Receiving one second phonetic entry, this second phonetic entry is carried out speech recognition to produce one second key word;

This second key word is carried out natural language processing to produce should second phonetic entry, one second user view; And

From this first repayment answer of first candidate list, select the second repayment answer according to this second user view.

2. the system of selection based on speech recognition as claimed in claim 1, wherein select the step of this first repayment answer to comprise according to this first user view:

The record and this first user view that are stored in a structured database are compared; And

When this record and this first user view are at least part of coupling, then this record is considered as this first repayment answer corresponding to this first phonetic entry.

3. the system of selection based on speech recognition as claimed in claim 1, wherein according to selecting the step of this second repayment answer to comprise in this first repayment answer from this first candidate list of this second user view:

Judge whether this second user view comprises an order vocabulary of indication order;

When this second user view comprises this order vocabulary of indication order, then in this first candidate list, select to be positioned at the first repayment answer of correspondence position according to this order vocabulary;

When this second user view does not comprise this order vocabulary of indication order, then will compare corresponding to record and this second user view of respectively this first repayment answer in this first candidate list; And

Which this first repayment answer that determines this first candidate list according to this comparison result is to should second phonetic entry.

4. the system of selection based on speech recognition as claimed in claim 3 wherein comprises step that should second phonetic entry according to which determines in this first candidate list this first repayment answer of this comparison result:

Select in this first repayment answer this matching degree to be should second phonetic entry for the soprano.

5. the system of selection based on speech recognition as claimed in claim 1, wherein the step of carrying out corresponding operation according to selected this first type of repaying answer comprises:

When selected this first type of repaying answer is music shelves, then music is carried out in selected this first repayment answer;

When selected first type of repaying answer is image shelves, then image is carried out in the selected first repayment answer and play;

When selected first type of repaying answer is webpage shelves, then the selected first repayment answer is shown;

When selected first type of repaying answer is picture shelves, then picture is carried out in the selected first repayment answer and show; And

When selected first type of repaying answer is a cardfile, then the selected first repayment answer is dialed and connected.

6. mobile terminal apparatus comprises:

One voice receiving unit receives one first phonetic entry and one second phonetic entry;

One display unit;

One storage unit is in order to store a plurality of data; And

One data processing unit, couple this voice receiving unit, this display unit and this storage unit, this data processing unit carries out speech recognition to produce one first key word to this first phonetic entry, this first key word is carried out natural language processing to produce one first user view that should first voice, and select the first repayment answer according to this first user view, when the quantity of this first repayment answer of selecting is 1, this data processing unit carries out corresponding operation according to the type of selected this first repayment answer, when the quantity of this first repayment answer of selecting greater than 1 the time, this data processing unit is controlled this display unit and is shown this first candidate list that comprises this first repayment answer, and this data processing unit carries out speech recognition to produce one second key word to these second voice, this second key word is carried out natural language processing producing one second user view that should second phonetic entry, and according to selecting the second repayment answer in this first repayment answer from this first candidate list of this second user view.

7. mobile terminal apparatus as claimed in claim 6, wherein this data processing unit will be compared corresponding to record and this first user view of respectively this first repayment answer, when respectively this record and this first user view are at least part of coupling, then this record is considered as this first repayment answer corresponding to this first phonetic entry.

8. mobile terminal apparatus as claimed in claim 6, wherein this data processing unit judges whether this second user view comprises an order vocabulary of indication order, when this second user view comprises this order vocabulary of indication order, then this data processing unit selects to be positioned at this first repayment answer of correspondence position in this first candidate list according to this order vocabulary, when this second user view does not comprise this order vocabulary of indication order, then this data processing unit with in this first candidate list respectively this first repayment corresponding this record of answer and this second user view compare to determine this first matching degree of repaying answer and this second phonetic entry, and this first repays answer corresponding to this second phonetic entry which to determine in this first candidate list according to these matching degrees.

9. mobile terminal apparatus as claimed in claim 8, wherein this data processing unit selects in this first repayment answer the matching degree soprano for to should second phonetic entry.

10. mobile terminal apparatus as claimed in claim 6, wherein the type when selected this first repayment answer is music shelves, then this data processing unit carries out music according to selected this first repayment answer, when selected this first type of repaying answer is image shelves, then this data processing unit carries out the image broadcast according to selected this first repayment answer, when selected this first type of repaying answer is webpage shelves, then this data processing unit shows according to selected this first repayment answer, when selected this first type of repaying answer is picture shelves, then this data processing unit carries out the picture demonstration according to selected this first repayment answer, and when selected this first the repayment answer type be a cardfile, then data processing unit according to selected this first the repayment answer dial and connect.

11. an infosystem comprises:

One servomechanism is in order to store a plurality of data and to have speech identifying function; And

A kind of mobile terminal apparatus comprises:

One voice receiving unit receives one first phonetic entry and one second phonetic entry;

One display unit;

One data processing unit, couple this voice receiving unit, this display unit and this servomechanism, this data processing unit carries out speech recognition to produce one first key word by this servomechanism to this first phonetic entry, first key word is carried out natural language processing to produce one first user view that should first phonetic entry, and this servomechanism is selected corresponding at least one first repayment answer and is sent to this data processing unit according to this first user view from the record that a structured database comprises, when the quantity of this first repayment answer of selecting is 1, this data processing unit carries out corresponding operation according to the type of selected this first repayment answer, when the quantity of this first repayment answer of selecting greater than 1 the time, this data processing unit is controlled this display unit according to this first repayment answer of selecting and is shown first candidate list that comprises this first repayment answer, and this data processing unit carries out speech recognition to produce one second key word by this servomechanism to this second phonetic entry, second key word is carried out natural language processing producing one second user view that should second phonetic entry, and this servomechanism is according to selecting the second repayment answer and be sent to this data processing unit in this first repayment answer from this first candidate list of this second user view.

12. infosystem as claimed in claim 11, wherein this servomechanism respectively compare by this record and this first user view of this first repayment answer, when respectively this record and this first user view are at least part of coupling, then this record is considered as corresponding this first repayment answer of this first phonetic entry.

13. infosystem as claimed in claim 11, wherein this servomechanism judges whether this second user view comprises an order vocabulary of indication order, when this second user view comprises this order vocabulary of indication order, then this servomechanism selects to be positioned at this first repayment answer of correspondence position in this first candidate list according to this order vocabulary, when this second user view does not comprise this order vocabulary of indication order, then this servomechanism with in this first candidate list respectively this record and this second user view compare to determine the matching degree of this first repayment answer material and this second phonetic entry, and according to this matching degree determine this first candidate list which this first repay answer corresponding to this second phonetic entry.

14. infosystem as claimed in claim 13, wherein this this servomechanism selects in this first repayment answer this matching degree soprano for to should second phonetic entry.

15. infosystem as claimed in claim 11, wherein the type when selected this first repayment answer is music shelves, then this data processing unit carries out music according to selected this first repayment answer, when selected this first type of repaying answer is image shelves, then this data processing unit carries out the image broadcast according to selected this first repayment answer, when selected this first type of repaying answer is webpage shelves, then this data processing unit shows according to selected this first repayment answer, when selected this first type of repaying answer is picture shelves, then this data processing unit carries out the picture demonstration according to selected this first repayment answer, and when selected this first the repayment answer type be a cardfile, then data processing unit according to selected this first the repayment answer dial and connect.

16. the system of selection based on speech recognition comprises:

One first phonetic entry is carried out speech recognition to produce one first key word;

Retrieve to obtain at least one first repayment answer according to this first key word in a structured database;

When the quantity of this first repayment answer of selecting greater than 1 the time, show that one comprises first candidate data of this first repayment answer;

After showing this first candidate list, receive one second phonetic entry, and this second phonetic entry is carried out speech recognition to produce one second key word; And

From this first repayment answer of first candidate list, select the second repayment answer according to this second user view.

17. the system of selection based on speech recognition as claimed in claim 16, wherein this first key word step of retrieving to obtain at least one first repayment answer in a structured database comprises:

When the record of this structured database and this first key word are at least part of coupling, then this record is considered as this first repayment answer corresponding to this first phonetic entry.

18. the system of selection based on speech recognition as claimed in claim 16, wherein according to selecting the step of this second repayment answer to comprise in this first repayment answer from this first candidate list of this second key word:

When this second key word comprises this order vocabulary of indication order, then in this first candidate list, select to be positioned at the first repayment answer of correspondence position according to this order vocabulary;

When this second key word does not comprise this order vocabulary of indication order, then will compare corresponding to record and this second key word of respectively this first repayment answer in this first candidate list; And

Which this first repayment answer that determines this first candidate list according to this comparison result is to should second phonetic entry.

19. the system of selection based on speech recognition as claimed in claim 18 wherein comprises step that should second phonetic entry according to which determines in this first candidate list this first repayment answer of this comparison result:

Select in this first repayment answer this matching degree to be should second phonetic entry for the soprano.

20. the system of selection based on speech recognition as claimed in claim 16, wherein the step of carrying out corresponding operation according to selected this first type of repaying answer comprises:

When selected this first type of repaying answer is music shelves, then music is carried out in selected this first repayment answer;

When selected first type of repaying answer is image shelves, then image is carried out in the selected first repayment answer and play;

When selected first type of repaying answer is webpage shelves, then the selected first repayment answer is shown;

When selected first type of repaying answer is picture shelves, then picture is carried out in the selected first repayment answer and show; And

When selected first type of repaying answer is a cardfile, then the selected first repayment answer is dialed and connected.

Images