Mamdami Fuzzy Inference System Research Papers

Mango Grading By Using Fuzzy Image Analysis

2023

In agricultural industry the efficiency and the proper grading process is very important to increase the productivity. Currently, the agriculture industry has a better improvement, particularly in terms of grading of fruits, but the... more

descriptionView Paper arrow_downwardDownload

Determining ‘Carabao’ Mango Ripeness Stages Using Three Image Processing Algorithms

by Cinmayii Manliguez

2023

Harvested mangoes are commonly classified or sorted manually. This method is tedious, time consuming, inaccurate and prone to errors. Human inspection is also subjective and factors like visual stress and tiredness may arise that can... more

descriptionView Paper arrow_downwardDownload

Metaphorical Paraphrase Generation: Feeding Metaphorical Language Models with Literal Texts

by Giorgio Ottolina

2022, Cornell University - arXiv

This study presents a new approach to metaphorical paraphrase generation by masking literal tokens of literal sentences and unmasking them with metaphorical language models. Unlike similar studies, the proposed algorithm does not only... more

descriptionView Paper arrow_downwardDownload

Mango Grading By Using Fuzzy Image Analysis

by DR. KHAIRUL ADILAH AHMAD

2022

In agricultural industry the efficiency and the proper grading process is very important to increase the productivity. Currently, the agriculture industry has a better improvement, particularly in terms of grading of fruits, but the... more

This study summarized that in this study, fuzzy image analysis can be applied to make mango grading in Matlab by combine the digital image processing and fuzzy classification. So this phase will discuss about the result that obtains from this study. Digital image processing in Matlab had been used actually is to extract the parameter or attribute of mango which is size, color and skin in order to prepare the input for fuzzy classification. Fig.2. The actual image of mango

The threshold does not need to be sensitive in this part. In fact, sometimes we face additional edges inside the object and affected the size negatively. So to remove these edges and keep the border, the ‘fill’ method is applied to get the correct size as shown in Table 1. For the color and skin process, the average from the three color RGB components is obtained and based on the number of edges we determined the skin image as shown in table 1.

descriptionView Paper arrow_downwardDownload

Mango Grading By Using Fuzzy Image Analysis

by Mahmod Othman

2022

In agricultural industry the efficiency and the proper grading process is very important to increase the productivity. Currently, the agriculture industry has a better improvement, particularly in terms of grading of fruits, but the... more

descriptionView Paper arrow_downwardDownload

Robust extraction of metaphors from novel data

by Laurie Feldman

2022

This article describes our novel approach to the automated detection and analysis of metaphors in text. We employ robust, quantitative language processing to implement a system prototype combined with sound social science methods for... more

descriptionView Paper arrow_downwardDownload

Metaphor Detection with Cross-Lingual Model Transfer

by Anatole Gershman

2022, Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

We show that it is possible to reliably discriminate whether a syntactic construction is meant literally or metaphorically using lexical semantic features of the words that participate in the construction. Our model is constructed using... more

descriptionView Paper arrow_downwardDownload

Predicting Concreteness and Imageability of Words Within and Across Languages via Word Embeddings

by Anita Peti-Stantic

2022, Proceedings of The Third Workshop on Representation Learning for NLP

The notions of concreteness and imageability, traditionally important in psycholinguistics, are gaining significance in semantic-oriented natural language processing tasks. In this paper we investigate the predictability of these two... more

descriptionView Paper arrow_downwardDownload

Fuzzy Ripening Mango Index Using RGB Colour Sensor Model

by TAJUL ROSLI RAZAK

2022, Researchers World

Currently, the mango ripeness classification is determined manually by human graders according to a particular procedure. This method is inconsistent and subjective in nature because each grader has different techniques. Thus, it affects... more

descriptionView Paper arrow_downwardDownload

Mango Grading By Using Fuzzy Image Analysis

by TAJUL ROSLI RAZAK

2022

In agricultural industry the efficiency and the proper grading process is very important to increase the productivity. Currently, the agriculture industry has a better improvement, particularly in terms of grading of fruits, but the... more

descriptionView Paper arrow_downwardDownload

Fuzzy Ripening Mango Index Using RGB Colour Sensor Model

by TAJUL ROSLI RAZAK

2022

Currently, the mango ripeness classification is determined manually by human graders according to a particular procedure. This method is inconsistent and subjective in nature because each grader has different techniques. Thus, it affects... more

descriptionView Paper arrow_downwardDownload

Machine Vision Based Techniques for Automatic Mango Fruit Sorting and Grading Based on Maturity Level and Size

by Dr Chandra Sekhar Nandi

2021, Smart Sensors, Measurement and Instrumentation

In recent years automatic vision based technology has become more potential and more important to many areas including agricultural fields and food industry. An automatic electronic vision based system for sorting and grading of fruit... more

Fig. 1 Proposed model of machine vision based automated mango fruit sorting and grading system

Fig. 2. Images of five varieties mango having different maturity level. Ist row: “KU”, 2nd row: “SO”, 3rd row :“LA” and 4th row: “HI” and Sth row: “AM”. Images are taken with an interval of 2 days, shown in (a) Ist colomn: raw (M1) (b) 2nd colomn: semimatured (M2) (c) 3rd colomn: matured (M3) (d) 4th colomn: over-matured (M4)

Fig. 3 Extracted still frames from the video image with an interval of 5 frames, (a) frame no.10, sample entering into the imaging chamber (b) frame no.15, sample near the middle of the imaging chamber (c) frame no. 20, sample crossing the middle position of imaging chamber (d) frame no. 25, sample going out from imaging chamber Though the images were taken in controlled environment under fixed illumination of light of 120 lux with the help of tungsten filament lamp, but there were some noises in the picture. To remove these noises a simple median filter found to provide reasonable good performance but it is computationally intensive. For that pseudo-median filter [19] was used in the work, as it is computationally simpler and possesses many of the properties of the median filter. This filtering process often helped to obtain smooth continuous boundary of the mango.

Fig. 4 Filtered images of Mango, a raw mango image along with the obtained contour, b binary image of the same mango after removing the small patches, also shows the different positions and their name as used in the work In order to find the boundary or the contour of the mango, a graph contour tracking method based on chain-code was adopted. This algorithm found to work reasonable fast, as the boundary of the mango is not so complex. The details of algorithm can be found in [20]. Here a brief description is given. The algorithm first detects every run at each row and records every single run’s serial number and the corresponding start- pixel coordinates and end-pixel coordinates are stored in a table named as “ABS”. Through this method, the run-length code of the image is obtained. Then, a 3 x 3 pixel box is adopted to detect the relationship between the objective pixel and its eight-connected surrounding pixels. All the runs are categorized into 5 classes, and then each run’s serial number along with the corresponding class label was recorded in another table named as “COD”. Then, the algorithm searches the ABS-table and COD-table sequentially. Based on the coordinates and class of each run recorded on the table, the starting pixel of the contour is recognized and the contour is successfully followed. And the chain code is generated while following the contour. An image of mango along with the traced boundary is shown in Fig. 4.

Fig. 5 Variation of R, G and B value along the longitudinal axis

Fig. 6 a Variation of average R of entire mango with maturity level(day), b variation of average R in apex region with maturity level(day), ¢ variation of difference of average R and G in equator with maturity level(day), d variation of average R to average G with four different maturity level(i.e. M1, M2, M3 and M4 of KU)

Fig. 7 a PDF distribution with average value of B and G of entire mango, b PDF distribution with the data set of mango collected from different batches and different gardens, and ¢ box-whiskers plots for most corelated features

Fig. 8 Flowchart of sorting and grading process

Fig. 9 Flowchart of Fuzzy Inference System

Fig. 11 Input membership functions representing maturity level (a) and size (b)

Fig. 12 Membership function representing quality output Table 2 Fuzzy if-then rules

Fig. 13 Defuzzification results from rule viewer

Fuzzy rule based algorithm is used to sort the mango fruit into four quality grades ike poor (Q1), medium (Q2), good (Q3) and very good (Q4) based on their maturity evel and size. The grading system has two inputs (maturity level and size) and one yutput (quality). Total 16 fuzzy if-then rules are created in order to sort the mango ruits into four different quality levels (Q1, Q2, Q3, and Q4). The fuzzy rule base s shown in Table 1. Examples of some rules are illustrated as in Table2. The fuzzy nference system is shown in Fig. 10. The rule viewer shown in Fig. 13 consists of the system’s inputs and output col-

Table 1 Fuzzy if-then rules for classification 8 Sorting and Grading Methodology Using Fuzzy Logic Technique

descriptionView Paper arrow_downwardDownload

High-Precision Abductive Mapping of Multilingual Metaphors

by Jerry Hobbs

2021, Proceedings of the Third Workshop on Metaphor in NLP

Metaphor is a cognitive phenomenon exhibited in language, where one conceptual domain (the target) is thought of in terms of another (the source). The first level of metaphor interpretation is the mapping of linguistic metaphors to pairs... more

descriptionView Paper arrow_downwardDownload

Review of Object Sorting Machine Using IOT

by subhash konar

2020, International Engineering Research Journal (IERJ),

This paper presents review of object sorting machine using Internet Of Things (IOT). Sorting is basically arranging the various objects according to their physical characteristics. Sorting is done in many large as well as small scale... more

COMPARATIVE STUDY OF VARIOUS SYSTEMS USED IN SORTING

descriptionView Paper arrow_downwardDownload

Mango Grading By Using Fuzzy Image Analysis

by Mahmod Othman

2016

In agricultural industry the efficiency and the proper grading process is very important to increase the productivity. Currently, the agriculture industry has a better improvement, particularly in terms of grading of fruits, but the... more

descriptionView Paper arrow_downwardDownload

A Corpus of Rich Metaphor Annotation

by Suzanne Wertheim

2015, Proceedings of the Third Workshop on Metaphor in NLP, NAACL 2015

Metaphor is a central phenomenon of language, and thus a central problem for natural language understanding. Previous work on the analysis of metaphors has identified which target concepts are being thought of and described in terms... more

Table 4: Metaphor candidates and their associated events

E. The comparison graph of “sheep” with translation “==” extended by HowNet’s sememes

Table 2: Impact of various axiom sources on span-selected Metaphor Mapping performance. Learned and Hybrid ap- proaches generally outperform the Manual approach to axiom collection.

Figure 2: Distribution of warrior metaphor centered around diagnosis event (x-axis: months from event, y- axis: average frequency of metaphor usage)

il Certain metaphor/event pairs show a peak around he event, or at 1 year after the event, for example on he anniversary of diagnosis, which is a significant event in the life of a cancer patient. However, the pattern does not hold across all such pairs, making it difficult to generalize. For example, in Figure 1, we see a peak of metaphor frequency occurring at he time of the event, but in Figure 2, we do not see such a peak at the time of the event, but see other peaks both before and after the event date. Another complicating factor is that different users experience different cancer treatment timelines. For instance, one user might experience these events over a long period of time, whereas another user may encounter these events in quick succession (Wen and Rosé, 2012). These factors motivated us to consider other methods, including hierarchical mixed models, for more in-depth analysis. Figure 1: Distribution of journey metaphor centered around diagnosis event (x-axis: months from event, y- axis: average frequency of metaphor usage)

Table 1: Corpus-wide unambiguous popular metaphor use statistics (among posts where the user used the metaphor at least once) (M: posts that contain each metaphor, L: posts that do not contain each metaphor).

Table 3: Metaphor use statistics of data used for MTurk (* indicates metaphor candidates for which the literal usage is more common than the non-literal one, N: nonliteral use L: literal use).

Table 2: Metaphor candidates and their associated events

The word 4f(pig) is in the center surrounded by its properties. Each property is linked to a bilingual sememe in HowNet (blue nodes).

The word 44(pig) is in the center surrounded by its properties. [he width of the line indicates the frenqucies of the “word-property” pair:

The 4#f(pig) has many properties(yellow nodes), and these yellow nodes have their words(orange nodes). So the extendted orange nodes share more properites with 44(pig) appear closer to the center.

Figure 1: Weights assigned to the different concreteness bins by the logistic regression classifier with the baseline feature set in an unweighted training regime. The bins span the 1-5 range with 0.25 increments; words falling in bin | are the most abstract, while words falling in bin 17 are the most concrete. Table 3: Performance of a model with optimally weighted training examples in comparison to the auto-weighted scheme, in terms of Precision (P), Recall (R), F-1 score (F), and the difference between Recall and Precision (D). A-B and B-A correspond to training-testing scenarios where the system is trained on Set A and tested on Set B and vice versa, respectively. All other figures report average performance across the cross-validation folds.

Table 5: Performance of a model trained with auto- weighted examples with and without DC (difference in concreteness) features.

Concreteness for literal (L) and nonliteral (NL) for heads vs. dependents, by P Figure |: Plot of concreteness scores for literal vs. nonliteral/metaphorical heads vs. their dependents, in the VUAMC grouped by parts of speech

Figure 2: Plot of imageability scores for literal vs. nonliteral/metaphorical heads vs. their dependents, in the VUAMC. grouped by parts of speech

Figure 3: Plot of anew scores for literal vs. nonliteral/metaphorical verbs in the VUAMC

Figure 1: Directionality of sensory modalities as pro- posed by Williams(1976).

Figure 2. Non-metaphoric frame network pattern The MetaNet project represents an effort to formal- ly represent and categorize metaphors and frames that comprise their source and target domains, and relations between them. Frames are coherent, con- ceptual gestalts organized in a hierarchical struc- ture. These range from experiential universal structures such as Motion Along a Path and Verti- cality, to more specific situations such as Physical Restraints and Disease; they also include less phys- ically concrete culturally-based frames like Pov- erty and Corruption. More-specific frames incorporate the semantics and internal structure of the more-general frames they inherit from, forming a complex network of related concepts. Relations between frames define how elements of a parent frame are incorporated by the child frame. For in- stance, the ‘subcase of’ relation indicates that the child fully inherits and elaborates the structure of the parent frame. In addition to traditional ontolog- ical relations, we also include relations specific to frame semantics and metaphor theory. A fragment of this network is illustrated in Figure 2.

Metaphors are essentially representations of mappings between frames. The structure of the source domain frame maps onto the structure of the target domain frame (Figure 3); hence, in POVERTY IS A DISEASE, the impoverished people of the Pov- erty frame are understood as the patient experienc- ing the disease in the Disease frame. Specifically, the roles of the Disease frame map onto their coun- terparts of the Poverty frame.

Figure 4. Structures accessed by phrase cure poverty.

Figure 6. Number of extracted metaphor expressions that evoke various translational motion frames

The richness of the extracted data and the struc- tural relations identified via our repository enable us to analyze data at varying levels of specificity. For instance, a search of the English Gigaword corpus (Graff & Cieri 2003) for metaphor expres- sions with the target domain ‘poverty’ revealed several interesting patterns. Firstly, at the very general level of frame families, we observe that the most frequently occurring source terms were either location/motion related (e.g. being at a location, translational motion, motion impediments) or in- volved physical harm (e.g. disease, physical com- bat, or other harmful encounters). Figure 5 shows the relative frequency of these frame families.

Figure 7. Number of different lexical units in extracted data that evoke the Downward Motion frame. Looking at a yet more specific level, we can ex- amine which lexical items are used to evoke a giv- en frame. Figure 7, below, shows this data for the Downward Motion frame. Looking at a yet more specific level, we can ex- amine which lexical items are used to evoke a giv- en frame. Figure 7, below, shows this data for the Downward Motion frame.

Figure 1: Abduction-based metaphor processing pipeline.

To determine predominant Conceptual Meta- phors for each protagonist (21% and 20% referred above), we rank order the Source Domains (SDs) for each side by number of LMs that use each SD. In Table 4, we show the predominant conceptual metaphors used for key targets by each protagonist. Table 4. The most representative CMs on both sides of the Gun Debate, by key Targets. Font size indicates relative frequency for top CMs for each target.

Figure 6. The INDO side dominates debate with use of more action-oriented language. Size of pie chart repre- sents the proportion of metaphors in the source role cat- those source domains that are used at least 2% of the time overall by both sides and the count of distinct rela- tions in the LMs by each protagonist. Y-axis represents count of metaphors.

Figure 4. The INDO side (red bars) dominates debate with use of more LMs overall. Here we show those source domains that are used at least 2% of the time overall by both sides and the count of LMs for those Source Domains. Y-axis represents count of metaphors. used in LMs that are of the “taking action” type (primarily verbs describing events, such as “at- tack’) vs. the “passively observe” (primarily nouns and adjectives, such as “devastating’’).

Figure 5. The INDO side dominates debate with richer vocabulary suggesting greater influence. Here we show a oe ah a La a. ae Ss. + ae oe,

Using this method, we find that the extreme proponents of the INDO and GOVTO sides are far apart, approximately 0.55 of the maximum theoret- ical distance of 1.0. Using the same measures, the distance between the INDO moderate position and both INDO and GOVTO extremes is approximate- y half of the above, or 0.27. This places the INDO moderate position in the center of the spectrum of positions between the two extremes. On the other hand, the language used by the GOVTO moderate position places them closer to the GOVTO ex- reme. This finding is illustrated in Figure 8.

Table 5: Performance on metaphor disambiguation evaluation. (6) is significantly better than (5) [p=0.013] (fs.: used feature selection)

Table 2: Performance of a model with AutoWeighted training examples in comparison to the unweighted base- line, in terms of Precision (P), Recall (R), and F-1 score (F) for the positive (’metaphor’”) class. A-B and B-A cor- respond to training-testing scenarios where the system is trained on Set A and tested on Set B and vice versa, re- spectively. All other figures report average performance across the cross-validation folds.

Table 4: Performance of a model trained with unweighted examples with and without DC (difference in concrete- ness) features.

Table 7: Performance the baseline model UPT+CBins in the baseline configuration (unweighted) the UPT+CUpDown+DCUpDown model in opti-weighted configuration.

Table 8: Instances of the adjective full in Set B that are predicted to be non-metaphors by the UPT model trained on Set A in the unweighted regime, while the UPT+CUpDown+DCUpDown model classifies these as metaphors. The noun that is recognized as being in the amod relation with full is shown in square brackets. FULL (small caps) indicates an instance that is annotated as a metaphor; lowercase version corresponds to a non- metaphor annotation.

Table 1: Percentages of POS and metaphors per register, with standardised residuals in brackets (n=47934

Table 4: Rules extracted from random forest models for Nouns (len=length of condition, freq=proportion of data instances meeting condition, err=proportion of incorrectly classified instances as over instances meeting condition, SD=standard deviation) Table 3: Results (Accuracy, F1) for random forest models for different POS, for predicting non/literal items, for models with and without the “weight” term w

Table 2: Results (F1) of evaluating models with different combinations of features, for predicting non/literal items (n=3574)

Table 5: Rules extracted from random forest models for Verbs (len=length of condition, freq=proportion of data instances meeting condition, err=proportion of incorrectly classified instances for instances meeting condition, SD=standard deviation, depslist=list of dependents)

Table 6: Rules extracted from random forest models for Prepositions (len=length of condition, freq=proportion of data instances meeting condition, err=proportion of incorrectly classified instances for instances meeting condition) concreteness, imageability, w, and a small but inter- esting subset of the ANEW categories, across POS. For example, while nouns make good use of con- creteness, im ageability and w, combinations of im- ageability and w are prevalent for verbs and prepo- sitions. Fur words, seem tures, but pre her, nouns and verbs, being content to make good use of the ANEW fea- positions make no use of such features, perhaps due to their status as function words. More careful study of a wider range of rules, as well as possible conditioning environments is required, and such suggestions remain tentative. Note also that one complication in all of this is that there are exten- sive errors for most of the extracted rules, and close study of possible sources of such errors is planned

Table 1: Feature sets used in the experiments.

Table 2: The cross validation and test accuracies of the baseline with and without vector space features.

Table 3: Performance of the B’ baseline in combination with the different sets of sensorial features. Statistical significance: ***, p < .001; **, p < .01; *, p < .05.

expression bear trap. Hence, disambiguating be- tween these two types of expressions requires a second step of metaphoricity evaluation.

specify document structural constraints, including grammatical constraints. To search texts for con- structional matches, we construct Resource De- scription Framework (RDF) models of each sentence in terms of an ontology defined in the Web Ontology Language (OWL). The ontology defines the classes Document, Sentence, and Word, and properties, some of which are shown in Table 2 and with their domain and range.

Table 3. Performance over gold standard data To evaluate the accuracy of the metaphor extractor, linguists collected attested sentences and annotated metaphoric expressions for the target domains Government, Bureaucracy, Democracy, Poverty, Taxation, and Wealth; they annotated all in- domain metaphoric expressions in the sentences where both the target and source were explicitly realized. Sentences were manually annotated for source and target word forms, source and target frames, and the construc tional pattern used to ex- press the metaphor. The metaphor extractor was run on these col ected go d standard sentences, and the output compared to the annotations entered by the linguists. Table 3 sho tions in the gold standard ws the number of annota- , the recall (percentage of gold standard examples that were identified), and the precision ( percentag e of extracted examples that were correct) of the system for three lan- guages.

Table 1: The number of metaphoric sentences in the devel- opment and test sets for each language. However, for a goal of high-precision mapping, sophisticated learning methods are not necessary. In- stead, we require that a chosen percent of the in- stances of a logical form fragment correspond to a single source or target concept, in which case we out- put a lexical axiom mapping the LF to the concept. We found it helpful to enforce the mutual exclusivity of the text fragments that map to source concepts and hose that map to target concepts. When a text frag- ment is ambiguous between a source mapping and a arget mapping, we produce an axiom for whichever correspondence was more frequent. This reduces the ikelihood of the axioms leading the system to iden- ify, e.g., two source concepts in a sentence but no arget concept. The results are axioms like

Table 2: Statistics for the corpus of annotations being released. Each category consists of one or more scenarios. For a variety of sentences, the corpus gives instantiations (slot—value pairs) of the slots in the schema for each scenario. Each linguistic metaphor is also tagged as being about one or more source and target concepts. All counts are for unique entries, except for slot—value pairs, which includes duplicates when they occur in the data. Some sentences contain more than one metaphor, so the number of unique sentences is less than the sum of unique sentences for each schema.

Table 1. Distribution of collected data across targets in Gun Case debate Step four is data collection from the sites closer to the middle of the opinion spectrum identified in the second step. When this data is processed for metaphors, they are labeled accordingly as “mod- erate”. We note that “moderate” positions in mul- ti-side conflicts may have different interpretations than in a largely binary conflict of Gun Case. In Table 1, the column Total Passages represents the sum total of passages processed from the extreme and moderate websites.

Table 3. Conceptual Metaphors used by protagonists on the GOVTO side Table 2. Conceptual Metaphors used by protagonists on the INDO side

To establish topical positioning, we first identify meso-topics that are present in a debate, as dis- cussed in Section 4.1. We then distinguish multiple forms in which polarization or valuation is applied to meso-topics in protagonists’ utterances such as through express advocacy or disadvocacy or via supporting or dissenting information, and express agreement or disagreement with a polarized state- ment made in a statement by the same or another protagonist. We create Topical Positioning Vectors representing each protagonist. Table 5 shows a fragment of positional vectors for extreme GOVTO and INDO positions for five meso-topics. In these vectors, value in each cell represents a prevailing combined polarity and intensity towards a meso-topic. We note that meso-topics form a su- perset of metaphorical targets as explained earlier. Table 5. Topical Positioning vectors for extreme GOVTO and INDO positions in the gun debate Topical Positioning vectors can now be used to calculate distance between protagonists, using standard cosine measure. We used this method to compute 4-ways distances in the Gun Case: be- tween the extreme positions on each side; between the moderate and extreme positions within each side; as well as between moderates and extremes across the sides and compared the distances so ob- tained to those obtained from metaphorical matri- ces (Section 5.3). We note that both methods

Table 2. Four Kinds of Word-Property Pairs But we find some important phenomena in the re- sults that will be introduced in section 4.2. 4 Statistics

Next we explore what the most frequent words and properties are in Chinese CogBank. This is important as we could learn what the most com- mon entities are that people tend to use as vehicles in similes, and the most common properties people prefer to express in everyday life. As shown in Ta- ble 4, nouns like flower, man, water, child, human, cat, angel, wolf and sunshine rank the highest in the database. These words are quite common in everyday life and they have hundreds of properties. But the top 10 properties of each word dominate more than half the occurrences of these words when employed in a simile. This indicates that people always rely more heavily on a word’s sali- ent properties to form a simile expression.

Table 5 shows the most frequent properties in CogBank: beautiful, bloom, fight, fly, convenient, warm, and painful. Each property is associated with hundreds of words. But the frequency of the top 10 concept words occupies more than half the occurrences. This indicates that people is tend to use the same kinds of vehicles to form a simile expression.

Table 5. Top 10 Most Frequent Properties in CogBank 4.2 Valuable Information from Uninforma- tive Data

The manual data check drops many uninformative data which on the surface seem to possess no value, for example, “as stupid as yow’’, “as cheap as be- fore’. The pronouns and time expressions have to be removed from CogBank. But through observing all the pronouns and time expressions through manual data check, we find something useful in Chinese sentences “X {& (as) Y —#¥ (same) A” (X is as A as Y) where Y is the reference object. As Indicated in Table 6, people prefer to use #¢(J) as the reference object rather than other pronouns.

People also prefer to reference recurring and concur- rent time frames over past or future ones. As shown in Table 7, usual (7£ 7%, #7) occurs more than past and before, while future(#X) occurs with even lower fre- quencies. q t S t J The usage patterns showing much higher fre- uencies for the pronoun #(/) and time expression tE i, Ff (usual) suggest that people prefer to use heir experienced everyday life knowledge to make imile or contrast sentences. This finding supports he Embodied Cognition Philosophy (Lakoff & ohnson 1980; Lakoff 2008), which hypothesizes t hat much of our conceptual structure is based on knowledge formed through physical and emotional experience.

descriptionView Paper arrow_downwardDownload

Grading of Carabao Mango Using Image Processing

by SDIWC Organization

2015

Grading the Quality of the mango is usuallyhandled physically which might not beaccurate due to the judgement of a person. Inthis study, image processing using MATLABis proposed using MandamiFuzzyInterference System. The Mango is placed... more

From the collected image, the computer program will collect the height and diameter of the sample mango by the use of boundary tracing. From the traced boundary, the program will measure the distance between one boundary to the other and gather the largest X and Y distance for the height and diameter respectively.

Figure 2. Spot Detection using Regionprops function The maturity will be based on the red means of the sample object. As the target colours are Green and Yellow, the only value that varies is the Red pixel values as a high red pixel value, a high green pixel value and a low pixel value produces a yellow colour and a low red pixel value, a high green pixel value and a low pixel value produces a green colour. The program will start by segmenting the colour between the mango and the background before collecting the

Figure 4.Membership function of Height variable

Figure 5. Membership function of Spot variable

A sample mango (Fig 8) which was 12 cm. tall was used to scale the distance between pixels as well as to check the limit of the height of the mango in pixels. From this mango sample, a 533 px tall mango is produced. Calculating for the limits of the acceptable height of an exportable mango, the for the lowest acceptable height of the mango, which is 10.5 cm, is scaled to 466 px; the highest acceptable height of the mango, which is 14 cm, is scaled to 622 px. Figure 8. An unripe mango and a ripe mango.

Figure 6.Membership function of Maturity variable There are five rules in the Fuzzy Inference system

Table 1 Input of the samples and the output from the Fuzzy Inference System The results show that the program is accurate enough to grade the mango. The system is strict enough to reject the sample mango if it is too large, is too small, is yellow or is blemished. While the system is more lenient on the grading between the spot detection considering our reference shows that there must be 0% tolerance in the spot detection of the system, it still show that the system may not be perfect as the REFERENCES

descriptionView Paper arrow_downwardDownload