Analysis Of Key Features Influencing Semantic Intensity Of Will And Would

With the context, even though there is an ambiguous word in a statement, people canpick out the right meaning of the word to make the statement understandable. However，some words which have complicated meanings cannot be easily defined in this way. It isthe problem that word sense disambiguation (WSD) dealt with. With time goes on, manyscholars in the field of linguistics realized WSD can be vey helpful in their studies. Theysuccessfully combined the corpus linguistics and artificial intelligence together. Based onthe statistics which is large corpus collected from the real world and according to differentmachine learning methods, the computer will reveal the deep knowledge of language.They have done great works in nouns and simple verbs but seldom touch other words.Modal verbs are complicated mainly because they have a tighter relationship with themood and emotion of human being. Even though modal verbs are indeterminate and themeaning of them may be difficult to be appropriately interpreted, they are frequently used.Therefore, building an effective and accurate WSD model for the English modal verbs hasgreat and significant meaning.The English modal verbs will and would are like twins who share many similarities intheir meanings. Now it is generally believed the meaning of modal verbs can be mainlydivided into two kinds: root meaning and epistemic meaning. According to Coates (1983),epistemic meaning is connected with predication or belief. Root meaning which is alsocalled non-epistemic meaning relates to volition, ability or obligation. As for the modalverbs will, root meaning is associated with volition and epistemic meaning relates topredication. Would shares same root meaning and epistemic meaning with will butpresents a weaker semantic intensity compared with will.To find the reason for their difference in semantic intensity, this research will startfrom the scope of modal verb WSD and try to build WSD model by adopting StructuralPartial-Ordered Attribute Diagram (SPOAD). Trough the analysis of WSD model, keyfeatures influencing semantic intensity of will and would will be found. First, based on aone million words corpus, this research extracts ten linguistic features, including eightsemantic features and two syntactic features from samples. By adopting SPOAD willWSD and would WSD model are built. After the test of “Leave-one-out” approach the WSD models are testified to be effective. Next, samples are divided into two partsaccording to their meaning. Fifteen semantic features and three syntactic features areextracted from samples which have root meaning. Sixteen semantic features and threesyntactic features are extracted from samples which have epistemic meaning. By adoptingSPOAD again two WSD models for semantic intensity are built. Finally, based upon theeffectiveness of WSD model and linguistic features, this research deletes semantic featuresor syntactic features to build new WSD models and testifies them. The experimental resultshows that semantic features contribute more to the WSD models of semantic intensity.Wherein, Mutual Information (MI) between will and its subject contributes the most insemantic intensity WSD model when will and would have root meaning, while MIbetween would and the verb after it has the greatest contribution to semantic intensityWSD model when will and would have epistemic meaning.Compared with WSD researches on a single modal verb, this research focuses on thecomparison between two English modal verbs will and would and explores from a moremeticulous scope and fill the gap of previous studies left behind. It will enrich and lead anew direction for the WSD researches on modal verbs.