Word Sense Disambiguation Of English Modal Verb May By Formal Concept Analysis

Ambiguity is the common phenomenon in natural language processing.Word sensedisambiguation is the alternative selection of meaning in certain context which has been a hotand tough issue in natural language processing and also become the key link in the fields ofmachine translation, information retrieval, text classification, speech recognition andhuman-machine interaction. In recent years, the achievements made in word sensedisambiguation were obviously witnessed, but the object of the study is still mainly focused oncommon nouns and verbs. The study of English modal verbs is relatively less compared withother studies. The combination with the FCA is seldom seen in natural language processing.Modality usually conveys the speaker’s opinion and attitude towards the proposition expressed,which is mainly realized by the using of modal verbs. Therefore, the establishment of an efficientand accurate model for word sense disambiguation is of great significance.The complexity and ambiguity of modal verb as well as its importance in humancommunication make the study of word sense disambiguation of great significance. This thesisattempts to establish the word sense disambiguation model of English modal verb May byFormal Concept Analysis proposed by Rudolf Wile. The WSD model was established byadopting the generation tool of attribute partial sequence diagram. This research, based on1.2million words corpus, extracts a series of semantic and syntactic features for the identification ofthe meaning of May. After that the software of Wconcord is used for the statistic of linguisticfeatures. Then the mutual information of the subject and may and that between verb and may arecalculated based on the statistic obtained through “Wconcord”. Different syntactic features wereselected,which may influence the meaning of “may”, and the logical value of “1”or “0” wasgiven for the presence or absence of the feature. The vectorized features constitute the formalcontext and then the WSD model of English modal verb may is established. The accuracyreaches92.6%.Based upon the established disambiguation model, the thesis also attempts to determine thecontributions of each feature to model formation. The contributions of different features to WSDare obtained through the deletion of one feature one time in sequence. It is also noted that thesubject has high classification towards the meaning of the root meaning granting permission. Thebinary disambiguation of only the root and epistemic meaning of may makes the contribution ofthe subject on WSD insignificant. However, this thesis tries to disambiguate the three meaningsof may, which makes the impact that the subject imposed upon seem much more significant owing to the participation of the authoritative organization or person in expression of the rootmeaning granting permission. The experiments also prove that the semantic features contributemore than the syntactic features to WSD.The establishment of English modal verb may not only be able to alleviate the burden of theresearchers to realize the automatic tagging of corpora but also can improve the accuracy ofmachine translation. Hence, the study result has theoretical and practical significance inlinguistic studies.