Chinese Verb Metaphor Recognition And Application Based On Semantic Knowledge

There are many non-literal expressions in natural language, such as metonymy, metaphor of words. Metaphor plays an important role in these abnormal expressions. Metaphor computing plays an important role in the fields of affective computing, text entailment, information retrieval and machine translation. Metaphor is not only a linguistic phenomenon but also a kind of a cognitive phenomenon.The research of metaphor computing at home and abroad is analyzed in this thesis, and metaphor recognition algorithms based on words similarity and word abstractness are proposed. Twenty common verbs are selected for metaphor recognition. The details are as follows:1. Verb metaphor recognition based on word similarity. In the metaphorical and literal usages of verb, this thesis observes that in the same usages of words, the similarity is large, while in the different usage of words, the similarity is small. Therefore, words metaphor recognition tasks is converted into words literal and metaphorical binary classification tasks. On the basis of existing semantic knowledge resources HowNet and Tongyici Cilin, metaphor recognition algorithms based on the similarity of the HowNet and the synonym information of Tongyici Cilin are brought up.2. Verb metaphor recognition based on word abstractness. In the binary classification of the verb, the literal class of verb is usually found to be visible, touchable, and concrete. On the contrary, the metaphor class is abstract concept which is more difficult to understand. According to this phenomenon, metaphor recognition algorithms based on abstract degree of word are given, which is calculated by HowNet and Tongyici Cilin.3. Verb metaphor recognition based of simple heuristic rules. To identify the verb sentence metaphor, the relative distance of the verb on both sides of the noun is used to determine the object of the verb, and the abstract degree of object is calculated to identify verb metaphor.4. Verb metaphor recognition based on dependency parsing. HIT IRLab Shared-Parser is used to parse the sentence. And the object of the verb is extracted for metaphor identification.5. The information retrieval system with metaphor processing function. In view of the important role of the metaphor computing in the natural language processing, the metaphor identification is applied in specific information retrieval field, and is built an information retrieval system with metaphor processing function.6. Knowledge based on verb metaphor. Metaphor knowledge database of twenty verbs which includes literal and metaphorical usages of word is built for future work.In this thesis, the research methods are mainly based on the semantic knowledge. Meanwhile, some data resources are accumulated in this thesis to support metaphor and related research.