汉语多义动词义项辨识模式 = A Pattern of Chinese Verbs Sense Recognition

Abstract

The paper address a pattern of Chinese verb sense recognition in context. Verb sense recognition is a procedure and mechanism to distinguish the sense of each polysemous verb in certain context. There are two levels in the pattern of Chinese verb sense recognition: the static level and the dynamic level. The study select 329 polysemous verbs of Chinese as a sample, and take the sense tagged Corpus of Chinese Teaching Materials as a database. In the static level, all the attributes of grammatical combination, semantic combination and collocation of each sense of the verbs, which are the distinctive features in verb sense recognition, is analyzed and described. With statistics of the their distribution in the table of the external attributes of verb sense, the importance degree of each distinctive features in distinguishing the verb senses in the static level is contrasted. In the dynamic level, some of the distinctive features can be omitted or substituted by pronoun or located on different place of a text and outside the minor sentence, on which a particular polysemous verb is located, therefore the verb sense has to be recognized in a larger range in text than in a minor sentence and sometimes it is hard because of the lack of distinctive features. With statistics of the features? distribution in the three context levels of sentence, paragraph and whole text in the corpus, the tendency of the distinctive features? position in each context levels is analyzed. On basis of this analysis, it becomes clear that a polysemous verb sense can be disambiguated with all types of distinctive features in the context: the grammatical features, arguments as semantic role and other verbs that are semantically and structurally related to the polysemous verb. Moreover, the importance degree of each distinctive features in distinguishing the verb senses in the dynamic level is contrasted and arranged. In addition, in consideration of the impact on automatic verb sense disambiguation from the Lexicographic division of verb sense, a method of measuring the distance between senses of a verb, which can be the reference criterion of dividing senses of a verb, is demonstrated. In conclusion, each type of distinctive features play a different role in the pattern of verb sense recognition. In the static level, the collocation, objective argument and semantic type of verb are the most important features. In the dynamic level, the leading features are the object-argument, collocation, complement type, subject-argument, and semantic type of verb. The study of verb sense recognition is meaningful and worthwhile in the area of lexicography and word sense disambiguation.