Chinese Verb Metaphor Recognition Based On Hownet And Verbnet

As a common phenomenon in human language, Metaphor is the central issue of language and thinking. Today, the machine metaphor recognition problem has become popular in the field of natural language processing. Currently, the most of studies on Chinese metaphor recognition focused on the identification of nouns. However, compared with other types of metaphor, verbal metaphors appear more frequency in the actual text. Researchers should pay more attentions to this issue. This paper try to translate the Chinese verbal metaphor problem into English verbal metaphor problem, using the perfect English knowledge base (Verbnet) to finish the Chinese verbal metaphor research. Since the scale of Chinese metaphor knowledge base is small, we propose this method to overcome its drawback.This paper analyzes the characteristics of HowNet, HIT and Wordnet, then propose a semantic mapping method among these three databases. First, we translate the senses in HIT into concepts sets based on HowNet. Second, according to the English information of concept, we could translate the concepts sets into English sets. Finally, we calculate the maximum value of the intersection of the English sets and synsets in Wordnet, then build the mapping between the concepts in HowNet and synsets in Wordnet. The results of the experiments show that the accuracy of the method achieve over90percent. It proves the feasibility of the mapping method.After get the mapping between HowNet and Wordnet, this paper proposes a Chinese verb metaphor identification method based on HowNet and Verbnet. The method uses the thematic roles and its selectional restrictions to check whether the subject and object are feasible. Finally we select20sentences from the corpus of "reader" to verify the proposed method. The result shows the proposed method achieves the expect accuracy and proofs the validity of the method.In summary, this paper proposed a method named Chinese verb metaphor recognition based on HowNet and Wordnet. It is conducive to the further study of Chinese metaphor recognition, also provides a theoretical significance and reference value for metaphor recognition.