Classification Rule Based Feature Selection Of English Modal Verb May

Natural Language is full of semantic indeterminacy,and a word can express different senses in different contexts.Modal verbs are especially complex semantic system because modal verbs are used to convey human feelings and attitudes.The semantic complexity of the modal verbs has been a hot and tough issue in both linguistic studies and natural language processing(NLP).The previous researches on the word sense disambiguation(WSD)of English modal verbs have revealed that the most important issue is to have the effective features besides adopting effective algorithms in order to achieve a high accuracy of WSD.Therefore,the feature selection for the WSD of English modal verb is carried out in this study.Based on a corpus of 1.1 million words,and taking the English modal verb may as the target word and its three main senses(root permission,root possibility and epistemic possibility)as sense categorization,the feature selection for the WSD of may is carried out.The research method adopted is the class-exclusive-feature filtration method with the attribute partial-ordered structure diagram(APOSD)approaches the supplement.A hundred and fifty sample sentences with sense-tagged may(50 for each sense)are randomly selected from the corpus as objects,and some semantic and syntactic features co-occurred with may are extracted as attributes.Based on that,and the formal context which expresses the relationship between objects and attributes is built up.Then,a computer program is used to filter the class exclusive features,and the APOSD is adopted to improve the result of WSD.Finally,the optimal feature set is obtained.The results of this study show that the features of the point mutual information(PMI)between modal verb may and its subject or predicate verb,the tense of the predicate verb,and interrogative sentence are more effective features for the WSD of may than other features.The results of this study provide new methods and ideas for the study of attribute feature optimization of other modal verbs and other parts of speech.