A Comparative Analysis Of The Formal Semantic Difference Of Prime English Modal Auxiliaries And Quasi-Modals In Light Of Rough Set Theory

Lexical items in natural language can convey different meanings in accordance with different situations of utterance.The senses of English modal auxiliaries are even more complex than those of other parts of speech in terms of senses as they express human feelings,attitudes and judgments,etc.English modal auxiliaries are categorized into prime English modal auxiliaries and quasi-modals.Although sharing some semantic meanings,prime English modal auxiliaries differ from quasi-modals in numerous aspects of semantic meanings.Traditionally,semantic comparisons of prime English modal auxiliaries and quasi-modals are conducted mainly from the perspective of linguistics,inadequate from formal semantics and computational linguistics.The rapid development of Natural Language Processing(NLP)and Machine Learning has enabled the semantic comparison of prime English modal auxiliaries and quasi-modals in a formalized and quantitative way.Moreover,the discovery of similarities and differences between these two types of English modal auxiliaries from the perspective of formal semantic features will provide valuable theoretical and practical references to the NLP of English modal auxiliaries.In this thesis,CAN and BE ABLE TO are chosen as examples for prime English modal auxiliaries and quasi-modals respectively.Based on a self-built corpus of one million words,this thesis studies the formal semantic differences of CAN and BE ABLE TO in light of Rough Set Theory.First,according to the distribution of CAN and BE ABLE TO in the corpus,300 and 130 sample sentences are selected for CAN and BE ABLE TO to build up decision tables which consist of attributes from semantic,syntactic and contextual features.Then,the attribute reduction and decision rule generation are conducted with the help of Rough Set Theory.Finally,on the basis of generated decision rules,Word Sense Disambiguation(WSD)models for CAN and BE ABLE TO are constructed for comparing the formal semantic differences between CAN and BE ABLE TO.The accuracy of generated decision rules for the WSD of CAN and BE ABLE TO reach to 82.67% and 81.54% respectively.Through the comparison,the formal semantic differences of CAN and BE ABLE TO are mainly analyzed in three aspects,namely,1)differences between decision rules for the WSD of CAN and BE ABLE TO;2)differences between important attributes for disambiguating the senses of CAN and BE ABLE TO;3)differences between important attribute combinations for disambiguating the senses of CAN and BE ABLE TO.The results of this thesis may contribute to the semantic comparison of prime English modal auxiliaries and quasi-modals as well as the feature selection in WSD and NLP.