Academia.eduAcademia.edu

Figure 25 – uploaded by Carlos Bobed

See full PDFdownloadDownload figure
Figure 7. Italian e-commerce turnover at the end of 2011

Figure 7 Italian e-commerce turnover at the end of 2011

Related Figures (77)

Figure 1. Logical architecture of WebODRA
Figure 1. Logical architecture of WebODRA
Figure 2. A sample forum developed using WebODRA framework
Figure 2. A sample forum developed using WebODRA framework
Figure 3. PHP handling user’s input.
Figure 3. PHP handling user’s input.
One usage location that was pointed out by Reviewer-5 was marked as Not in use. The framework was not used there since it required a modification to the framework that would change its semantics, a task that the team preferred not to do at the time that that specific location in the application was implemented.
One usage location that was pointed out by Reviewer-5 was marked as Not in use. The framework was not used there since it required a modification to the framework that would change its semantics, a task that the team preferred not to do at the time that that specific location in the application was implemented.
capability to OSGi in order for OSGi to be more applicable to enterprise systems. Toward this goal, in our previous paper [13] we proposed an approach, where distribution is enabled by exposing OSGi services as Web services, which is done by creating a “middleware” bundle that adapts OSGi services to become Web services. Figure | gives an overview of our approach. Figure 1. Architecture of our approach [13]. (System B is an external system.)
capability to OSGi in order for OSGi to be more applicable to enterprise systems. Toward this goal, in our previous paper [13] we proposed an approach, where distribution is enabled by exposing OSGi services as Web services, which is done by creating a “middleware” bundle that adapts OSGi services to become Web services. Figure | gives an overview of our approach. Figure 1. Architecture of our approach [13]. (System B is an external system.)
Figure 19. GUI of airline interface. The airline interface lets airlines to change information about scheduled flights. For example, an airline can change the aircraft type to a smaller one if the currently scheduled plane cannot be filled or the airline can change the departure time if the currently scheduled plane is delayed due to weather conditions. All these changes are broadcasted to the flight list display system and the mobile client by the management system. Figure 19 shows GUI of the airline interface. The flight list display system sends updated flight information (including current departure time, gate, aircraft type and flight status: on-time, delayed or cancelled) to flight list displays, which are spread over the airport. Figure 20 shows GUI of the flight list display system.
Figure 19. GUI of airline interface. The airline interface lets airlines to change information about scheduled flights. For example, an airline can change the aircraft type to a smaller one if the currently scheduled plane cannot be filled or the airline can change the departure time if the currently scheduled plane is delayed due to weather conditions. All these changes are broadcasted to the flight list display system and the mobile client by the management system. Figure 19 shows GUI of the airline interface. The flight list display system sends updated flight information (including current departure time, gate, aircraft type and flight status: on-time, delayed or cancelled) to flight list displays, which are spread over the airport. Figure 20 shows GUI of the flight list display system.
To prove the feasibility of our approach, we implemented a flight information system (FIS) using our approach. The purpose of the FIS is to inform passengers waiting at the airport about flight changes made by airlines. Figure 18 gives an overview of the FIS. Figure 18. Architecture of flight information system.
To prove the feasibility of our approach, we implemented a flight information system (FIS) using our approach. The purpose of the FIS is to inform passengers waiting at the airport about flight changes made by airlines. Figure 18 gives an overview of the FIS. Figure 18. Architecture of flight information system.
Figure 21. GUI of mobile client. In addition, the mobile client lets passengers receive updated flight information from the comfort of their mobile phones. Figure 21 shows GUI of the mobile client.
Figure 21. GUI of mobile client. In addition, the mobile client lets passengers receive updated flight information from the comfort of their mobile phones. Figure 21 shows GUI of the mobile client.
Figure 1. Relational to Semantic Mapping Model With Relational data on one side and RDF repository data on the other, the mapping algorithms have involved domain specific heuristic formulation to satisfy the accurate implementation of knowledge transfer. Different reasoning logic and rules are part of the mapping algorithms.
Figure 1. Relational to Semantic Mapping Model With Relational data on one side and RDF repository data on the other, the mapping algorithms have involved domain specific heuristic formulation to satisfy the accurate implementation of knowledge transfer. Different reasoning logic and rules are part of the mapping algorithms.
Figure 2. Relational Database Schema Overview
Figure 2. Relational Database Schema Overview
Figure 13. RDF Schema view (Protége) The implementation of the algorithms is based on the Space Project Management scenario. An interactive Java application is used to execute the mapping procedures and the resulting Semantic Web repository loaded in Protégé (a free open-source Java tool providing an _ extensible architecture for the creation of customized knowledge-based applications) OWL editor is shown on the figure below. The Ontology diagram modeled above is also serialisable using OWL file format as in Figure 14 below.
Figure 13. RDF Schema view (Protége) The implementation of the algorithms is based on the Space Project Management scenario. An interactive Java application is used to execute the mapping procedures and the resulting Semantic Web repository loaded in Protégé (a free open-source Java tool providing an _ extensible architecture for the creation of customized knowledge-based applications) OWL editor is shown on the figure below. The Ontology diagram modeled above is also serialisable using OWL file format as in Figure 14 below.
BLE I. SUMMARY TABLE OF RDB-TO-RDF MAPPING LANGAGE COMPARISON [6] WITH RD2SW
BLE I. SUMMARY TABLE OF RDB-TO-RDF MAPPING LANGAGE COMPARISON [6] WITH RD2SW
RDB-TO-RDF MAPPING LANGUAGES COMPARISON LEGEND [6]
RDB-TO-RDF MAPPING LANGUAGES COMPARISON LEGEND [6]
Figure 15. Relational Database Schema search result The mapping test between the Relational Database Schema (Figure 2) and RDF Schema (Figure 13 & Figure 14) is evaluated using meta-data search comparisons as shown by the sample screenshots in Figure 15 & Figure 16 below.
Figure 15. Relational Database Schema search result The mapping test between the Relational Database Schema (Figure 2) and RDF Schema (Figure 13 & Figure 14) is evaluated using meta-data search comparisons as shown by the sample screenshots in Figure 15 & Figure 16 below.
Figure 16. RDF Triples search result We have compared our attempt to some of the existing approaches (Direct Mapping, eD2R, R20, Relational.OWL, Virtuso, D2RQ, Triplify, R2RML, R3M). We used the comparison criteria specified in the seminal paper [6]. The results of this comparison can be summarized as shown in Table I and Table II.
Figure 16. RDF Triples search result We have compared our attempt to some of the existing approaches (Direct Mapping, eD2R, R20, Relational.OWL, Virtuso, D2RQ, Triplify, R2RML, R3M). We used the comparison criteria specified in the seminal paper [6]. The results of this comparison can be summarized as shown in Table I and Table II.
Figure 1. A Basic Service Oriented Architecture [11]
Figure 1. A Basic Service Oriented Architecture [11]
Figure 3. The proposed Architecture with the Protocol service
Figure 3. The proposed Architecture with the Protocol service
the Protocol service executes the target service and ends the process. Alternatively, if the request involves two-way operations (synchronous), the Protocol service executes the target service and the result is eventually returned to the consumer. Figure 4. The outline of the Transformation method
the Protocol service executes the target service and ends the process. Alternatively, if the request involves two-way operations (synchronous), the Protocol service executes the target service and the result is eventually returned to the consumer. Figure 4. The outline of the Transformation method
Figure 8. E-shopping Scenario
Figure 8. E-shopping Scenario
Figure 2. Top 10 reviews sites by numbers of review
Figure 2. Top 10 reviews sites by numbers of review
Figure 3. Top 10 internet countries in Europe. The best way to get information about the holidays destinations is to take recommendations from friends, colleagues or relatives. “Sharing” is a key word of new way intending tourism; in fact reviews and social networks are connected with principal sources chosen by travelers for get information on travels. At the second place we can see Internet websites, at the third the personal experience. Other sources for reaching information are: travel/tourist agencies and free catalogues and brochures are important. Less important than the other are: newspapers, radio and TV, paid-for guidebooks and magazines and social media sites (see Fig.4). [13]. Fig. 5 shows the most important methods used for arranging the holidays. These are the results of several inquests, with multiple responses, of the document "Attitudes of European Towards Tourism", published by the European Commission, where analyzes the behavior of European tourists in 2012 [13]. As we can see, the main method used is Internet with a 53%, almost the double of the second method most used. These data can bring us to reflect about the importance of Internet and the experience (ours experience or people who we know) in comparison with other sources. From these results, it is easy to see the importance of the ICTs and web services in the tourist sector.
Figure 3. Top 10 internet countries in Europe. The best way to get information about the holidays destinations is to take recommendations from friends, colleagues or relatives. “Sharing” is a key word of new way intending tourism; in fact reviews and social networks are connected with principal sources chosen by travelers for get information on travels. At the second place we can see Internet websites, at the third the personal experience. Other sources for reaching information are: travel/tourist agencies and free catalogues and brochures are important. Less important than the other are: newspapers, radio and TV, paid-for guidebooks and magazines and social media sites (see Fig.4). [13]. Fig. 5 shows the most important methods used for arranging the holidays. These are the results of several inquests, with multiple responses, of the document "Attitudes of European Towards Tourism", published by the European Commission, where analyzes the behavior of European tourists in 2012 [13]. As we can see, the main method used is Internet with a 53%, almost the double of the second method most used. These data can bring us to reflect about the importance of Internet and the experience (ours experience or people who we know) in comparison with other sources. From these results, it is easy to see the importance of the ICTs and web services in the tourist sector.
Figure 4. Information sources considered to make a decision about travel plans (multiple responses)
Figure 4. Information sources considered to make a decision about travel plans (multiple responses)
In 2012 we can see in which products the Spanish people decide to use internet as a commerce vehicle use by 27% of
In 2012 we can see in which products the Spanish people decide to use internet as a commerce vehicle use by 27% of
Figure 6. Ten activities with the biggest turnover in the global electronic commerce in the first 2012 trimester
Figure 6. Ten activities with the biggest turnover in the global electronic commerce in the first 2012 trimester
Figure 9. Percentage of Internet Users in function of the total population
Figure 9. Percentage of Internet Users in function of the total population
This sector is used mainly by men (5,7%) instead women (2,7%) [26]. In 2012, the 56% of internet users (in France) used internet in order to travel. In other countries as Germany or U.K, this value is higher than the 64%, with a value of 71%, for U.K. Figure 10. Main information sources to make decision about travel destination.
This sector is used mainly by men (5,7%) instead women (2,7%) [26]. In 2012, the 56% of internet users (in France) used internet in order to travel. In other countries as Germany or U.K, this value is higher than the 64%, with a value of 71%, for U.K. Figure 10. Main information sources to make decision about travel destination.
RELATIONSHIPS BETWEEN TERMS its pattern rule, the matching logic will return a matching score because these two parameters are closely related (perhaps using the rules “CityName propertyOf City” and “codeOfCityv propertvOf City’). The semantic matching technique estimates the similarity of the input and output by considering the underlying concepts the input/output parameters cover. Formally, we describe the input as a vector 1 =< p;,C; > (similarly, t O =< py, Cy he output can be represented in the form >), where pj; is the set of input parameters and C; is the concept that is associated with p;. Then, the simila- rity of the in put can be found using the following two steps (the output can be processed in a similar fashion); (1) we split pj; into a set of terms, we then find synonyms for these terms, and (2) we replace each term with its corresponding concepts, and then compute a similarity score.
RELATIONSHIPS BETWEEN TERMS its pattern rule, the matching logic will return a matching score because these two parameters are closely related (perhaps using the rules “CityName propertyOf City” and “codeOfCityv propertvOf City’). The semantic matching technique estimates the similarity of the input and output by considering the underlying concepts the input/output parameters cover. Formally, we describe the input as a vector 1 =< p;,C; > (similarly, t O =< py, Cy he output can be represented in the form >), where pj; is the set of input parameters and C; is the concept that is associated with p;. Then, the simila- rity of the in put can be found using the following two steps (the output can be processed in a similar fashion); (1) we split pj; into a set of terms, we then find synonyms for these terms, and (2) we replace each term with its corresponding concepts, and then compute a similarity score.
TABLE II. EXAMPLE FOR DISCOVERY PROBLEM We describe an automatic Web API discovery algorithm similar to the one in [5]. An API matches a query when an API is sufficiently similar to the query. This means that we need to allow the agent to perform matches that recognize the degree of similarity between APIs and the query. We define the matching criteria as follows: matches the query requirement is the Web API discovery problem. For example, we are looking for an API to search a hotel. Table 2 shows the input/output parameters of a query and an API. In this example a Web API W satisfy the query Q. Q requires HotelName as the output and W produces HotelName and ConfirmNumber. The extra output produ- ced can be ignored. W requires CountryCode and Name- OfCity as the input and Q provides CountryID, State- Name, and CityName as the input. An API parameter can be matched with the other parameter only if there is a semantic relationship between them. Here, although CountryCode and CountryID are different forms, they have the same semantics since they are referred to the same concept. Also NameOfCity and CityName have the same semantics since they are properties of the same object (i.e., City). Therefore the agent is able to infer that Q and W input parameters have semantically the same classes.
TABLE II. EXAMPLE FOR DISCOVERY PROBLEM We describe an automatic Web API discovery algorithm similar to the one in [5]. An API matches a query when an API is sufficiently similar to the query. This means that we need to allow the agent to perform matches that recognize the degree of similarity between APIs and the query. We define the matching criteria as follows: matches the query requirement is the Web API discovery problem. For example, we are looking for an API to search a hotel. Table 2 shows the input/output parameters of a query and an API. In this example a Web API W satisfy the query Q. Q requires HotelName as the output and W produces HotelName and ConfirmNumber. The extra output produ- ced can be ignored. W requires CountryCode and Name- OfCity as the input and Q provides CountryID, State- Name, and CityName as the input. An API parameter can be matched with the other parameter only if there is a semantic relationship between them. Here, although CountryCode and CountryID are different forms, they have the same semantics since they are referred to the same concept. Also NameOfCity and CityName have the same semantics since they are properties of the same object (i.e., City). Therefore the agent is able to infer that Q and W input parameters have semantically the same classes.
Figure 1. General Overview of Query and Matching APIs Searching sub-graphs: First, we search the API registry about any API that has all the output parameters of the query (we call “last nodes”), and any API that has at least one of the input parameters of nodes”). After this searching it i sets are obtained for the first and the query (we call “first s assumed that non empty last nodes. The next is to create n-ary trees for every last node by visiting all the nodes connected to a particular last node. Such tree is constructed by recursively includ the directed similarity graph unti We use the BFS algorithm to so ing nodes and edges from 1 we reach the first nodes. lve this problem. Now we can find all the possible composition candidates from the trees. Figure 1 shows a general overview of the query and the matching APIs before constructing the overall composi- tion plans.
Figure 1. General Overview of Query and Matching APIs Searching sub-graphs: First, we search the API registry about any API that has all the output parameters of the query (we call “last nodes”), and any API that has at least one of the input parameters of nodes”). After this searching it i sets are obtained for the first and the query (we call “first s assumed that non empty last nodes. The next is to create n-ary trees for every last node by visiting all the nodes connected to a particular last node. Such tree is constructed by recursively includ the directed similarity graph unti We use the BFS algorithm to so ing nodes and edges from 1 we reach the first nodes. lve this problem. Now we can find all the possible composition candidates from the trees. Figure 1 shows a general overview of the query and the matching APIs before constructing the overall composi- tion plans.
Figure 3. Directed Similarity Graph To experiment with the data mashup tool we extracted a collection of REST and SOAP APIs from Programmable- Web. To avoid potential bias, we chose different APIs from different domains. We first collected a subset which associ- ated REST APIs for three domains: weather, travel, and mapping. This set contains 63 APIs. Next, we collected a subset containing 17 SOAP APIs from three domains: zip- code, location, and search. In Figure 3, we show a directed similarity graph which obtained from our experimental dataset. The graph consists of 80 nodes and 123 edges. A possible query for the Web API composition is given as follows: Q.1 = {zipcode}, Q.0 = {city, latitude, longitude}. The composition result is exemplified by part of the directed similarity graph as shown in Figure 4. From the registry our engine has discovered 8 last nodes (dark grey circles) and 7 first nodes (light grey circles). We call the BFS algorithm and create an n-ary tree for each last node. This is repeated until all the last nodes are reached.
Figure 3. Directed Similarity Graph To experiment with the data mashup tool we extracted a collection of REST and SOAP APIs from Programmable- Web. To avoid potential bias, we chose different APIs from different domains. We first collected a subset which associ- ated REST APIs for three domains: weather, travel, and mapping. This set contains 63 APIs. Next, we collected a subset containing 17 SOAP APIs from three domains: zip- code, location, and search. In Figure 3, we show a directed similarity graph which obtained from our experimental dataset. The graph consists of 80 nodes and 123 edges. A possible query for the Web API composition is given as follows: Q.1 = {zipcode}, Q.0 = {city, latitude, longitude}. The composition result is exemplified by part of the directed similarity graph as shown in Figure 4. From the registry our engine has discovered 8 last nodes (dark grey circles) and 7 first nodes (light grey circles). We call the BFS algorithm and create an n-ary tree for each last node. This is repeated until all the last nodes are reached.
visually, and iteratively refine their goals until the final results satisfying. The ontology learning method automati- cally builds semantic ontologies from Web API descriptions
visually, and iteratively refine their goals until the final results satisfying. The ontology learning method automati- cally builds semantic ontologies from Web API descriptions
ded results in Table 4 have desired or relevant goals. Although the 3rd ranking result turns out to be invalid as it does not satisfy the user requirement, top 2 ranking results have desired composition plans. These results have shown that our algorithm can generate most user desired outputs. TABLE IV. LIST OF RANKED COMPOSITION CANDIDATES
ded results in Table 4 have desired or relevant goals. Although the 3rd ranking result turns out to be invalid as it does not satisfy the user requirement, top 2 ranking results have desired composition plans. These results have shown that our algorithm can generate most user desired outputs. TABLE IV. LIST OF RANKED COMPOSITION CANDIDATES
Fig. 1. X3D interfaces and metadata nodes In the example in Listing 1, the shape is described by three elements: geometry, appearance and metadata. The whole semantics is enclosed in the Met adat aSet and includes both simple (subject, creator, description—lines 9-10) as well as complex (Mass, 11) nested nodes. Every node is assigned a name and a reference to the name definition (a web document). The containerField associates the node with the appropriate field of its parent node.
Fig. 1. X3D interfaces and metadata nodes In the example in Listing 1, the shape is described by three elements: geometry, appearance and metadata. The whole semantics is enclosed in the Met adat aSet and includes both simple (subject, creator, description—lines 9-10) as well as complex (Mass, 11) nested nodes. Every node is assigned a name and a reference to the name definition (a web document). The containerField associates the node with the appropriate field of its parent node.
Fig. 2. Mapping between RDFa attributes and X3D metadata nodes
Fig. 2. Mapping between RDFa attributes and X3D metadata nodes
Fig. |. | Technology made communication evolve from a one-to-one to a many-to-many model. The future could bring human-machine intercommunication.
Fig. |. | Technology made communication evolve from a one-to-one to a many-to-many model. The future could bring human-machine intercommunication.
In Table 1 some candidate terms are classified according to the Labour Market ontology classes.
In Table 1 some candidate terms are classified according to the Labour Market ontology classes.
Figure 1. Classes and instances of Labour Market ontology All those terms derive from the occupation descriptions and belong to the group of specific terms with the highest TF-IDF values (see point 3 in the previous Section B). Ontology classes and relationships have to be refined from the initial classification taxonomy of candidate terms. By taking as main references the ILO and O*Net resources it is possible to specify some ontology parts, e.g., Sector class with its sub-classes and instances as represented in Figure 1.
Figure 1. Classes and instances of Labour Market ontology All those terms derive from the occupation descriptions and belong to the group of specific terms with the highest TF-IDF values (see point 3 in the previous Section B). Ontology classes and relationships have to be refined from the initial classification taxonomy of candidate terms. By taking as main references the ILO and O*Net resources it is possible to specify some ontology parts, e.g., Sector class with its sub-classes and instances as represented in Figure 1.
TABLE III. HERO COMPETENCY QUESTIONS’ ANSWERS (EXCERPT)
TABLE III. HERO COMPETENCY QUESTIONS’ ANSWERS (EXCERPT)
TABLE II. SOME EXAMPLES OF HERO CQS ACCORDING TO THEIR RESPECTIVE CATEGORIES
TABLE II. SOME EXAMPLES OF HERO CQS ACCORDING TO THEIR RESPECTIVE CATEGORIES
Figure 1. Competency Questions Translation Approach
Figure 1. Competency Questions Translation Approach
We are conscious that our work encompasses several limitations, such as:
We are conscious that our work encompasses several limitations, such as:
Previously [4] we have introduced a model of stigmergy including the concept of a stigmergy grand purpose and the core components of stigmergy: the agent, the environment, and the sign. The model is illustrated in Figure 1. The model as illustrated in Figure 1. ties together the core components of stigmergy, an inner band representing tl attributes of the com the dynamics acting outer band dynamics or defining the inter feedback (contributi on those attributes. Fur are either internal to eac ons) and negative feed illustrating how the agent, where the negative feedback is ap environment. The paper where we presented t focused on a holistic model of stigmergy which applies for the world of entomology, the human world and the virtual world. This paper is focused specifically on how the varieties of stigmergy manifest as Web environment UI elements. In context to this paper, these three components correlate to the users of Web environments, contributions that the ponents, and an outer band representing thermore, t Our mod back (deca positive feedback is contributed by t plied by t users make. he model w Nn component, face between components. describes the dynamics of equilibrium between positive ne ne el y) he ne as and the
Previously [4] we have introduced a model of stigmergy including the concept of a stigmergy grand purpose and the core components of stigmergy: the agent, the environment, and the sign. The model is illustrated in Figure 1. The model as illustrated in Figure 1. ties together the core components of stigmergy, an inner band representing tl attributes of the com the dynamics acting outer band dynamics or defining the inter feedback (contributi on those attributes. Fur are either internal to eac ons) and negative feed illustrating how the agent, where the negative feedback is ap environment. The paper where we presented t focused on a holistic model of stigmergy which applies for the world of entomology, the human world and the virtual world. This paper is focused specifically on how the varieties of stigmergy manifest as Web environment UI elements. In context to this paper, these three components correlate to the users of Web environments, contributions that the ponents, and an outer band representing thermore, t Our mod back (deca positive feedback is contributed by t plied by t users make. he model w Nn component, face between components. describes the dynamics of equilibrium between positive ne ne el y) he ne as and the
Figure 2. Example of a Facebook LIKE mechanism.
Figure 2. Example of a Facebook LIKE mechanism.
Figure 3. Example of a Facebook SEEN-BY mechanism.. Another variety of stigmergy describes the development of unintentional trails within the environment. An example of this is best shown by people wearing a path into a lawn when a short-cut is taken across the lawn. This unintentional trail is similar to another type of UI mechanism found within Facebook (see Figure 3. ).
Figure 3. Example of a Facebook SEEN-BY mechanism.. Another variety of stigmergy describes the development of unintentional trails within the environment. An example of this is best shown by people wearing a path into a lawn when a short-cut is taken across the lawn. This unintentional trail is similar to another type of UI mechanism found within Facebook (see Figure 3. ).
ERaUI project uses case-based reasoning (CBR) paradigm [20, 21] to learn about the user’s profile and in developing the ERaUI adaptive web user interface. CBR uses machine algorithms to solve new problems by adapting solutions of previous similar problems following the cycle Retrieve, Reuse, Revise and Retain, as shown in Figure 2.
ERaUI project uses case-based reasoning (CBR) paradigm [20, 21] to learn about the user’s profile and in developing the ERaUI adaptive web user interface. CBR uses machine algorithms to solve new problems by adapting solutions of previous similar problems following the cycle Retrieve, Reuse, Revise and Retain, as shown in Figure 2.
We conducted an experiment within the university research community on the impact of usability methodology of personas and scenarios that led to the design of the ERaUI interface as an intelligent and portable widget that we deployed on the left hand side of NaCTeM web interface as shown in Figure 1.
We conducted an experiment within the university research community on the impact of usability methodology of personas and scenarios that led to the design of the ERaUI interface as an intelligent and portable widget that we deployed on the left hand side of NaCTeM web interface as shown in Figure 1.
Figure 3. ERaUI Click & Mouse Movement Heat Map. ERaUI use click / mouse movement heat maps techniques to tracks the user interaction with the host website as shown in Figure 3.
Figure 3. ERaUI Click & Mouse Movement Heat Map. ERaUI use click / mouse movement heat maps techniques to tracks the user interaction with the host website as shown in Figure 3.
Figure 4. User Records and brpwsing patterns
Figure 4. User Records and brpwsing patterns
Figure 6. Results of content filtering method ERaUI implements advanced filtering algorithm(s) for extracting digital content of the host web site like NaCTeM to match user needs. Content and collaborative based filtering methods [27, 23] have been used as the basis for selection and learning about the content of NaCTeM web site. Content methods suggest topics similar to the ones a user group with similar profile has liked in the past as shown in Figure 6:
Figure 6. Results of content filtering method ERaUI implements advanced filtering algorithm(s) for extracting digital content of the host web site like NaCTeM to match user needs. Content and collaborative based filtering methods [27, 23] have been used as the basis for selection and learning about the content of NaCTeM web site. Content methods suggest topics similar to the ones a user group with similar profile has liked in the past as shown in Figure 6:
Figure 9. ERaUI live help box As shown in Figure 9, ERaUI interface also provides live communication facilities through ‘live help' box enabling the user to seek support and guidance from the NaCTeM host web site administrator.
Figure 9. ERaUI live help box As shown in Figure 9, ERaUI interface also provides live communication facilities through ‘live help' box enabling the user to seek support and guidance from the NaCTeM host web site administrator.
Figure 8. Suggesting useful links for a category of users Also, based on the identified web user profile, ERaUI recommends choices on some aspects of the research processes as desirable or undesirable, rating them on a scale and giving some similar form of evaluation to help the user in his/her selection of retrieved information. An instance of ERaUI user interface adapting itself to a category of users is shown in Figure 8. This screenshot suggests to the user currently browsing the host web site a few links in yellow colour, normally used by users with similar web user profiles.
Figure 8. Suggesting useful links for a category of users Also, based on the identified web user profile, ERaUI recommends choices on some aspects of the research processes as desirable or undesirable, rating them on a scale and giving some similar form of evaluation to help the user in his/her selection of retrieved information. An instance of ERaUI user interface adapting itself to a category of users is shown in Figure 8. This screenshot suggests to the user currently browsing the host web site a few links in yellow colour, normally used by users with similar web user profiles.
Figure 7. Panes of similar pages from collaborative filtering
Figure 7. Panes of similar pages from collaborative filtering
Following the completion of a project, the materials contained in some of these resources are considered for archival purposes so that they remain available despite the end of funding and dispersal of teams. This is an increasingly familiar situation as funders and institutions seek to develop repositories to increase the impact and availability of their work. Such archives or repositories of web resources enable users to search for information in the collection using basic keyword searching which proves to be ineffective simply because they retrieve web pages, datasets, and research papers merely on the basis of incidental mentions of a term in users’ queries. Such a searching strategy misses the context, relationship, historical details and other attributes of a particular resource which would have enabled users to extend their exploration to other related records in the archive. One such web resources repository is the ReStore repository [6] which will be used as a test-bed for the experimentation and search performance evaluation described here. Figure 1: An overview of web resource, creation, development and its archival into a web resource archive, e.g., ReStore repository
Following the completion of a project, the materials contained in some of these resources are considered for archival purposes so that they remain available despite the end of funding and dispersal of teams. This is an increasingly familiar situation as funders and institutions seek to develop repositories to increase the impact and availability of their work. Such archives or repositories of web resources enable users to search for information in the collection using basic keyword searching which proves to be ineffective simply because they retrieve web pages, datasets, and research papers merely on the basis of incidental mentions of a term in users’ queries. Such a searching strategy misses the context, relationship, historical details and other attributes of a particular resource which would have enabled users to extend their exploration to other related records in the archive. One such web resources repository is the ReStore repository [6] which will be used as a test-bed for the experimentation and search performance evaluation described here. Figure 1: An overview of web resource, creation, development and its archival into a web resource archive, e.g., ReStore repository
Figure 3: ReStore synthetic ontology architecture to be integrated into the KIM’s bootstrap ontology KIMO to aid in (a) annotating domain-specific (Social Science Research Methods) resources in ReStore repository (b) searching relevant information in integrated ontologies. In Figure 3, we have put together very basic terms and concepts (classes, properties) in a container with a view to mapping and extending them to other ontologies and storing the resulting annotation metadata (instances of classes or KB together in the form of integrated ontologies. The KB and mapped network of ontologies are assumed to be searched at a time to produce relevant and accurate results with a view to evolving it iteratively as new terms are added. GATE will process structured/semi-structured content in documents and OWLIM represents RDF triple store containing inference, querying engine for information retrieval.
Figure 3: ReStore synthetic ontology architecture to be integrated into the KIM’s bootstrap ontology KIMO to aid in (a) annotating domain-specific (Social Science Research Methods) resources in ReStore repository (b) searching relevant information in integrated ontologies. In Figure 3, we have put together very basic terms and concepts (classes, properties) in a container with a view to mapping and extending them to other ontologies and storing the resulting annotation metadata (instances of classes or KB together in the form of integrated ontologies. The KB and mapped network of ontologies are assumed to be searched at a time to produce relevant and accurate results with a view to evolving it iteratively as new terms are added. GATE will process structured/semi-structured content in documents and OWLIM represents RDF triple store containing inference, querying engine for information retrieval.
The Cronbach’s alpha was calculated for each of the factors obtained in order to estimate the reliability of the results. Factor 1 obtained a Cronbach alpha value of 0.874, Factor 2 obtained a Cronbach alpha value of 0.907 and Factor 3 obtained a Cronbach alpha value of 0.862. This signifies that the values obtained are good, and that the results are therefore reliable.
The Cronbach’s alpha was calculated for each of the factors obtained in order to estimate the reliability of the results. Factor 1 obtained a Cronbach alpha value of 0.874, Factor 2 obtained a Cronbach alpha value of 0.907 and Factor 3 obtained a Cronbach alpha value of 0.862. This signifies that the values obtained are good, and that the results are therefore reliable.
TABLE III. FACTORIAL ANALYSIS
TABLE III. FACTORIAL ANALYSIS
RELATIONSHIP BETWEEN DEMOGRAPHIC ASPECTS AND CLUSTERS
RELATIONSHIP BETWEEN DEMOGRAPHIC ASPECTS AND CLUSTERS
TABLE IV. CLUSTER ANALYSIS
TABLE IV. CLUSTER ANALYSIS
E VI. SUMMARY OF THE RELATIONSHIP BETWEEN DEMOGRAPHIC ASPECTS AND CLUSTER Limitations of the study. This work has been carried out in a systematic manner. Nevertheless, we are conscious that it has certain limitations. The first concerns the quality model used since we have limited the research to DQ characteristics in the Contextual category of the SPDQM model.
E VI. SUMMARY OF THE RELATIONSHIP BETWEEN DEMOGRAPHIC ASPECTS AND CLUSTER Limitations of the study. This work has been carried out in a systematic manner. Nevertheless, we are conscious that it has certain limitations. The first concerns the quality model used since we have limited the research to DQ characteristics in the Contextual category of the SPDQM model.
Figure 1. Events that are similar by vocabulary but different. edge represents the relation between two minimal events. In order to establish EGM, we first define two types of a minimal event which can not be divided on a sentence in the given article: an event noun phrase and an event predicate-argument structure. An event noun phrase is a noun phrase which can represent an event by itself. If we divide the noun phrase into several elements, then no elements can represent an event. Therefore, it is adequate that we regard an event noun phrase as a minimal event. A predicate-argument structure in a sentence, abbreviated to PAS, is called an event predicate-argument structure, abbreviated to EPAS, if its predicate represents an action related to an event. For example, sentence “Date Muratoki revived the economy of his feudal domain” contains an EPAS because its predicate “revived” represents an action in the economic revival; on the other hand, the predicate of sentence “Tokugawa Ieyasu had been told as a descendant of Nitta Yoshishige” is not regarded as an action of a historical event. If a sentence has two predicates representing actions of events, then we regard the sentence contains two EPAS. We then introduce the following eight kinds of a relation between minimal events: deployment, illustration, parataxis, causality, paradox, progress, means, and situation. In contrast to models dealing with the causality relation between events [1], our eight kinds of relations are expected to find similar events more accurately. We also present an idea how to determine which events have a relation by utilizing the centering theory proposed by Grosz et al. [5], [6]. Figure 2 illustrates an example of composite events in
Figure 1. Events that are similar by vocabulary but different. edge represents the relation between two minimal events. In order to establish EGM, we first define two types of a minimal event which can not be divided on a sentence in the given article: an event noun phrase and an event predicate-argument structure. An event noun phrase is a noun phrase which can represent an event by itself. If we divide the noun phrase into several elements, then no elements can represent an event. Therefore, it is adequate that we regard an event noun phrase as a minimal event. A predicate-argument structure in a sentence, abbreviated to PAS, is called an event predicate-argument structure, abbreviated to EPAS, if its predicate represents an action related to an event. For example, sentence “Date Muratoki revived the economy of his feudal domain” contains an EPAS because its predicate “revived” represents an action in the economic revival; on the other hand, the predicate of sentence “Tokugawa Ieyasu had been told as a descendant of Nitta Yoshishige” is not regarded as an action of a historical event. If a sentence has two predicates representing actions of events, then we regard the sentence contains two EPAS. We then introduce the following eight kinds of a relation between minimal events: deployment, illustration, parataxis, causality, paradox, progress, means, and situation. In contrast to models dealing with the causality relation between events [1], our eight kinds of relations are expected to find similar events more accurately. We also present an idea how to determine which events have a relation by utilizing the centering theory proposed by Grosz et al. [5], [6]. Figure 2 illustrates an example of composite events in
Figure 2. Examples of minimal and composite events.
Figure 2. Examples of minimal and composite events.
EXAMPLES OF A PREDICATE STRUCTURE OF EVENTS. Table I
EXAMPLES OF A PREDICATE STRUCTURE OF EVENTS. Table I
Table III RESULTS OF APPLING OUR RULES OF PREDICATES. THE NUMBER OF RELATIONS BETWEEN MINIMAL EVENTS. Table II
Table III RESULTS OF APPLING OUR RULES OF PREDICATES. THE NUMBER OF RELATIONS BETWEEN MINIMAL EVENTS. Table II
PRECISION OF PREDICATE SEARCH BY EACH METHODS.
PRECISION OF PREDICATE SEARCH BY EACH METHODS.
RECALL OF PREDICATE SEARCH BY EACH METHODS.
RECALL OF PREDICATE SEARCH BY EACH METHODS.
a priest and renamed; (b) He succeed to a house because of retirement of his father; (c) He is helped to the lord, and returned to secular life; (d) He killed his men and he lost trust. We use three simple methods for search of these queries: AND search of two predicates; search wit! the information of predicates and the edge of the predicates in our event graph model; search with the information of predicates, the edge and its label in our event grap n n model. For this experiment, we use 500 Wikipedia articles of Japanese feudal lords. We also use the Japanese Wordnet [9] for comparison of two predicates.
a priest and renamed; (b) He succeed to a house because of retirement of his father; (c) He is helped to the lord, and returned to secular life; (d) He killed his men and he lost trust. We use three simple methods for search of these queries: AND search of two predicates; search wit! the information of predicates and the edge of the predicates in our event graph model; search with the information of predicates, the edge and its label in our event grap n n model. For this experiment, we use 500 Wikipedia articles of Japanese feudal lords. We also use the Japanese Wordnet [9] for comparison of two predicates.
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved