Chinese Verb Metaphor Recognition Based on Topic Model

Zhen-kai BAI,Xiao-xi HUANG,Rong-bo WANG,Zhi-qun CHEN,Xiao-hua WANG
DOI: https://doi.org/10.3969/j.issn.1673-629X.2016.11.015
2016-01-01
Abstract:Metaphor is an integral part of human language,and the quality of metaphor processing will directly affect the effectiveness of natural language processing and machine translation. Metaphor recognition is an essential task in metaphor processing as a foundational work and has got the attention of the researchers. At present,most Chinese metaphors recognition has focused on identifying the phrase level of noun metaphor,however,verbal metaphors has higher frequency in the actual text,which should be paid attention by more Chi-nese metaphor researchers. In order to improve the recognition rate of Chinese metaphor,in view of the Chinese verb metaphor,an ap-proach to metaphor recognition is proposed based on topic model. In this method,sentence topic distribution generated through LDA mod-el is used as a feature,and the metaphor recognition is implemented with SVM. The average accuracy of the method is 76. 46%,after fur-ther joined the feature of topic annotation,the average accuracy of the method is 80. 42%. The experimental results show that the method is effective.
What problem does this paper attempt to address?