Word Sense Disambiguation Of Quasi-modal-verb Be Able To And Co-occurrence Restriction Between Modal Verbs And Be Able To

Quasi-modal is a grouping formally distinguishable form,but semantically similar to the modal verbs in modern English,such as be able to,be going to,have to,have got to,need to.Compared with modal verbs,quasi-modals are often neglected and result in relatively rare.However,this does not match with their growing frequency in use.Word sense disambiguation(WSD)is one of key and difficult projects in natural language processing(NLP).WSD aims at homonyms such as adjectives,prepositions,modal verbs and so on,by principles like context information,syntactic features to intelligently select the most appropriate sense with the help of computer.Quasi-modal be able is selected as research object to investigate the knowledge discovery based on WSD in this dissertation.This study is based on the theory of Formal Concept Analysis that has been applied to the study of many modal verbs WSD with great achievement.The study designed in the paper WSD of quasi-modal be able to is on the basis of 2.5 billion corpus with the help of structural partial-ordered attribute diagram(SPOAD).Firstly,the senses of be able to are identified as “ability”,“possibility” and “permission”.Then,after building the training set with 215 samples,a WSD model of be able to is constructed,and its self-checking accuracy is 88.37%.At the same time,the accuracy counted in testing set with 110 samples is 94.55%.At last,the semantic and syntactic features influencing the senses of be able to are studied by rules extracted and analysis on attribute features.It is found that the three senses of be able to are all influenced by the subject and verb collocated with them;they all have relations with the syntactic features;compared with semantic features,syntactic features have less influence on the senses.Moreover,rules in the same sense are similar,while rules in different senses tend to be disparate.Based on the WSD model of be able to,knowledge discovery research is conducted.It is found that primary modal verbs are more frequently collocated with be able to than other quasi-modal verbs.By analyzing the most frequent sense combinations and zero frequency sense combinations,rules of sense combination of modal co-occurrence are generalized as: from the first modal(M1)to the second modal(M2),the two sensesexperience a change from strong to weak in term of subjectivity,that is,senses with higher subjectivity are ranked front.In addition,the structure of modal co-occurrence has the function of expressing the speaker's complicated emotion that not only involves subjective attitude but also concerns objective factors,which can not be replaced by single modal expression.Finally,as for the restrictive relations between the two modal verbs,M1 tends to be subjective,while the M2 prefers to be objective;the more subjectivity the sense involves,the higher the frequency of modal co-occurrence is.The research on quasi-modals not only widens the scope of semantic modality study,but also has shed light to the studies in natural language processing.For further study,this research has a guiding significance.