Word Sense Disambiguation And Rule Extraction Of Secondary English Modal Verb Could

Ambiguity is a common phenomenon that widely exists in daily communication.Identifying the specific meaning of an ambiguous word in a certain context is of virtualimportance in understanding a whole passage. Word sense disambiguation (WSD) is oneof the most challenging issues in natural language processing. It has been applied invarious fields, such as machine translation, information retrieval, text mining, textcategorization, speech recognition and human-machine interaction. Modality clearlyreveals a speaker’s attitude and intention. Therefore, word sense disambiguation of thecomplicated modal verbs plays a significant role in the study of human languages.This thesis attempts to establish the word sense disambiguation model of thesecondary English modal verb could by formal concept analysis (FCA) proposed byRodolf Wille. This paper builds a corpus of three milion words and divides the senses ofcould into three root meanings and one epistemic meaning based on Coates’ senseclassification of English modal verbs. It calculates the mutual information of the subject ofthe sample sentences and the four senses of could, and the mutual information of couldand the related verbs respectively. It extracts eight syntactic features of could in the realcontext in the corpus. Based on the method of formal concept analysis, this thesisconstructs the word sense disambiguation model of could. Its accuracy reaches92.33%,which proves the effectiveness of the formal concept analysis as a means of sensedisambiguation of modal verbs. Based on the model, the rules of word sensedisambiguation of the secondary English modal verb could are extracted. The accuracyreaches92%. Moreover, this paper uses another way to extract WSD rules. Based on thesimplified formal context, the author induces the attribute features of the modal verb could.The experiment shows that the unique attribute of the congeneric objects and the uniquecomposite attributes of the congeneric objects play significant roles in the sensecategorization of could. The re-examination accuracy of the extracted rules reaches94%.Based on the results, this thesis makes a further comparison and induction of the two kindsof methods of rule extraction. This paper disambiguates the senses of the secondary English modal verb could withthe theory and method of Formal Concept Analysis and extract the rules of Word SenseDisambiguation on the basis of the generated attribute positive sequence diagram.Moreover, this thesis adopts the attribute features method to extract rules and achievesrelatively high accuracy. The current research broadens the horizon of natural languageword sense disambiguation research and provides theoretical and practical references forthe semantic study of the secondary English modal verbs and natural language processing.