TAMIL THESAURUS TO TAMIL WORDNET

TAMIL THESAURUS TO TAMIL WORDNET Rajendran Sankaravelayuthan Amrita Vishwa Vidyapeetham , Coimbatore rajushush@gmail.com Abstract Tamil Onto-thesaurus is an outcome of a very long research activity that went on in the field of lexical semantics of Tamil vocabulary. It went through several stages before being culminated into Tamil ontothesaurus. It depicts our travel from Tamil thesaurus to Tamil word net. It is a lexical resource which amalgamates all sorts of information available in a dictionary, thesaurus and wordNet. The Dravidian wordNets (in which Tamil wordNet is one of the four components) built under the IndoWordNet project depends on an ontology developed by western conceptualization of vocabularies of a language i.e. English). This has not taken into consideration the Indian conceptualization of vocabulary depicted in the nikhandu tradition. Say for examples, nikhandus have classifications such a six types of tastes, nine types of planets (gragams), 7 types of mandalams (a type of division), 15 tidis (15 phases of moon), etc which are crucial for Indian tradition. In the western oriented WordNet ontology there is no scope for the visualization of concepts depicted in nikhandus. More over building a wordNet based on Hindi wordNet which in turn is built on English wordNet will take many years to complete and it would miss the conceptualization depicted in Indian tradition. The present onto-thesaurus is based on the Indian conceptualization of vocabulary and the process of building one is very simple. We have the plan to make it into a generic one so that all the Dravidian languages can be easily accommodated into it. 1. Introduction A paper thesaurus for Tamil was prepared in 1990 based on the principles of componential analysis of meaning propounded by Nida (1975) and was published in 2001 (Rajendran, 2001), nearly after a decade. Following the paper thesaurus, an Electronic thesaurus for Tamil was attempted and a book on Tamil electronic thesaurus was published in 2006 (Rajendran and Baskaran, 2006) The preparation of wordNet for Tamil was undertaken (2001-2003) with the financial assistance from Tamil Virtual University (renamed now as Tamil virtual academy) and a crude version of it based on the ontology developed by Rajendran (Rajendran, 2001) was submitted to the institute in 2003. After that, from 2009 onwards with the fund received from MHRD and Department of electronics and information Technology of Govt. of India the building of Dravidian wordNet was executed based on Hindi wordNet; nearly 3000 synsets (concepts) have been completed. Still we have to go a long way to achieve the desired target. At present a team from CEN, Amrita University is involved in building onto-thesaurus for Tamil as a part of the project e titled Co puti g Tools for Ta il La guage teachi g a d lear i g . The project is funded by Tamil Virtual Academy, Chennai. 2. Tamil Onto-thesaurus Thesaurus is a in its wider sense is a classification of words by concepts, topics, or subjects. The present Tamil Onto-thesaurus is the extended version of Electronic thesaurus of Tamil focusing more on the ontological features. Two kinds of issues arise in the preparation of Tamil onto-thesaurus:   Linguistic issues Computational issues 2.1. Linguistic Issues It involves mainly the following four tasks: 1. Developing an ontology for Tamil based on structural semantic principles. 2. Establishing semantic domains and sub domains based on distinguishing semantic or Componential features of lexical items. 3. Classifying Tamil vocabulary to fit into the ontology developed. 4. Linking words by various semantic or lexical relations such as synonymy, hyponymyhyperonymy, meronymy-holonymy, compatibility, and incompatibiliity. 2.2. Computational Issues It involves mainly the following three tasks: 1. Conversion of linguistic data base into computer accessible format. 2. Preparation of a tool to provide the facilities for augmenting, entering and editing the raw data, and classifying the lexical items in a semi-automatic way. 3. Creation of user friendly interfaces for accessing the onto-thesaurus in simple manner. 3. Ontology of Tamil vocabulary The ontology available in Rajendran (2001), which is founded on the theory of componential analysis of meaning propounded by Nida (1975) is enhanced to suit the present purpose. The following is the skeletal structure of the Tamil ontology adopted in Onto-thesaurus. 1. எ. 1. எ. . . 2. எ. . : . . 1. , , 2. 3. 2. , , , , , , ; , , , , 3. , , , - , , , - , , , எ , , , , , , . 4. , , , . Nida (1975:15) advocates for four principal ways in which meanings of different semantic units can be related to one another. They are inclusion, overlapping, complementation and contiguity. These relations help us to classify the vocabulary of Tamil in terms of different semantic domains and distribute the lexical items in a structural fashion under these domains. 4. Structuring of vocabulary by lexical relations Lexical semantics offers foundation for structuring vocabulary in terms of lexical relations (Lyons 230335, Cruses 1986). In the NLP oriented papers, the general practice is to avoid giving linguistic details based on which the system is built. But here we would like to give the lexical semantics of building ontothesaurus to make it more transparent. 4.1. Congruence Relations The four basic relations between classes furnish a model for establishing fundamental group of sense relations and for defining a set of systematic variants applicable to virtually all other paradigmatic sense relations: Identity, inclusion, overlapping, and Disjunction (Cruse, 1986:86-87). Identity: class A and class B have same members. Inclusion: class B is wholly included in the class A Overlap: class A and class B have members in common but each has members not found in the other disjunction: class A and class B have no members in common These four congruence relations culminates into the four lexical relations discussed below. 4.2. Lexical Relations There are at least four lexical or meaning relations by which lexical items can be linked or related to one another in the ontological structure of Tamil vocabulary. They are synonymy, hyponymy, compatibility and incompatibility. A word acquires its referential meaning in being a member of a semantic domain by the common features it shares with other members in that domain, and having contrasting features which separate it from other members of the domain. It is the semantic relations among words, such as synonymy, hyponymy, compatibility and incompatibility (Cruse, 1986:84-111), which help one to classify and organize words in terms of semantic domains in a structural fashion. ): Synonymy ( e.g. - ( 'pet': 'animal’; meronymy-holonymy ‘cow’: : e.g. ): 'book'; hyponymy-hypronymy 'book': ; Compatibility ( : ‘hill’: ): e.g. 'dog'; Incompatibility ( ): e.g. ‘water hole’. Incompatibility leads to the relation called opposition which culminates into many types which are discussed below. 4.3. Lexical Inheritance Hyperonymy-hyponymy and meronymy-holonymy assure lexical item to inherit semantic features as exemplified below: – . - இ , . – . – . 4.4. Opposition There are many types of oppositions (Lyons, 1977: 270-290, Cruse 197-263). They are listed below with examples. எ Gradable opposites ( ): e.g. எ opposites ( , : எ , ): : , ; Ungradable : ; Complementaries e.g. : : ; Privative opposition ( : ): One denotes some positive property and the other denotes the absence of that property, e.g. எ ; Equipollent opposition ( ]: The contrasting lexemes denotes a positive propert e.g. ; Converseness ( : Converseness is distinguished from antonymy and complementarity e.g, , : ’ எ ; 'X Y- 1. Converse pairs of social roles: e.g. : Converse pairs of Kinship terms e.g. / relations: e.g. : : : , , ; ,எ : / ): , : ’. /எ : ; 2. ; 3. Converse pairs of Temporal ; 4. Converse pairs of Spatial relations: e.g. : Directional opposites: e.g. opposites: e.g. : 'Y X- : ): : : , , ; 'arrive': : , ; : ; Orthogonal ; Antipodal opposites: : : Non-binary oppositions If the opposition involves more than two lexical items such contrasts are called non-binary opposition (Lyon, 1977:271). There a number of types of binary contrasts. They are dealt under non-branching hierarchies. 4.5. Hierarchies Hierarchies are of two types: branching hierarchies and non-branching hierarchies. The branching hierarchy shows tree structure, where as non-branching hierarchy does not show tree structure (Cruse 1986:112). Branching hierarchies Non-branching hierarchies A P B D C E F Q G R S 4.5.1. Taxonomic hierarchies The hierarchies are the outcome of the hypronymy-hyponymy relation between lexical items. The sequence of hyperonymy-hyponymy relation leads the hierarchical structure of a set of vocabulary items which show this pair of relations among themselves. Taxonmic hierarchies are more liberal than hypronymy-hyponymy hierarchy. The following is an example: எ 4.5.2. Meronymic hierarchies Meronymic hierarchies are the result of meronymy-holonymy relation shown by the lexical items. The meronymy-holonymy relation also gives hierarchical structure to a set of vocabulary. The following is an example. 4.5.3. Non-branching hierarchies Non-binary opposition leads to a number of types of non-branching hierarchies. They are listed below: Bipoles The bipoles (Cruse, 2000: 189) is the simplest kind of linear structure found in a pair of opposites. They are simply oppositions which we have discussed earlier. Bipolar chains The bipolar chains (Cruse, 2000: 189) have implicit superlative terms of opposite polarity at each end of the scale. The following is the example: , , , , Monopolar Chains In monopolar chains (Cruse , 2000: 190) there is no sense that the terms at the ends of the chain are oriented in opposite directions. Degree, stages, measures, ranks and sequences come under monopolar chains. Degrees: Degrees (Cruse, 2000: 190) incorporate as part of their meaning different degrees of some continuously scaled property such as size or intensity. e.g. : : : Stages: Stages (Cruse 2000: 190) are points in a life cycle of something and normally involve the notion of progression. e.g. , , , , ; ; , , , , Measures: Measures (Cruse, 2000: 190) are based on part-whole relationship, with each whole divided into a number of identical parts. e.g. , , , ; , , , , , , , , ; , ; , , , Ranks: : The lexical items under ranks entails a sequential order which is not gradual (Cruse, 2000: 191): e.g. , , , Sequence: The lexical items which are ordered but does not bear increasing property are sequences (Cruse, 2000: 191). e.g. , , , …, , , , எ , , , ; , Cyclical sets or cycles: The sequential lexical item can entail a cyclical order of time in the natural arena (Cruse 1986:187-190). e.g. { , , , , , }; { , Propositional series or grids , , , } , , , , , …}; { , The units of a gird is the cell, which consists of four lexical items, any one of which must be uniquely predictable form the remaining three (Cruse, 1986:118-131; Cruse, 2000: 191-193) . The followings are the examples of cells: e.g. :: : :: : ; : :: : ; : : 5.User friendly interface for accessing the onto-thesaurus A few user friendly interfacs have been prepared from which one access the infomration needed form the onto-thesaurus of Tamil. 5.1. Linguistic Tree Viewer in Java NLP or Linguistic researchers who work in syntax often want to visualize parse trees or create linguistic trees for analyzing the structure of a language. TreeViewer software provides an easy to use interface to visualize or create simple linguistic trees. This software is written entirely in Java. This tree viewer has been converted to depict the ontological structure of Tamil vocabulary. The semantic relations such as synonymy, hyponymy-hypernymy, meronymy-holonymy, oppositions, entailments are captured by the tree viewer. The tree representation will be converted into an ontology based visual thesaurus. 5.2 Tree structure of a domain As a sample for demo the semantic domain nava kirangkaL 'nine planets' has been taken and converted into tree structure as shown in the snap-shot given below: (( ( ( , ( , , ( , , , ( , , , , , ; ))))) ) , , )); , , , , ; , ( , ( , , )) , ; , , (எ ) , , , , ; , ; ; ; There is one more GUI which gives the hierarchical details of a lexical items there by one will be able to get the meaning of the lexical item form the hierarchical information. This shown in the following screen shot. 6. Conclusion The coverage of the vocabulary at present is only 30000 lexical items. We hope to improve on it in the near future. We like to accommodate all kinds of lexical and meaing relations or linkages a user expects from the Tamil onto-thesaurs. All the information availabel to a word and a set of words will be incorporated in the onto-thesarus of Tamil. The present onto-thesurus system will be converted into a generic system so as to accommodate all the other Dravidian languaes. The onto-thesaurus under preparation has wide range of uses which inclue information retrieval across Dravidian languages, machine translation across dravidian languages and building knowledge systems for Dravidian languages. References (1) . . 2001. . : . (2) , . & , . . 2006. : . (3) Nida, E.A. 1975. Componential Analysis of Meaning: An Introduction to Semantic Structure. The Hague: Mouton. (4) Cruse, D.A. 1986. Lexical semantics. Cambridge: Cambridge University Press. (5) Cruse, D. A. Meaning in Language: An Introduction to semantics and pragmatics. Oxford: Oxford University Press. Lyons, J. 1977. Semantics, volume 1, Cambridge: Cambridge University Press.