摘要
针对现有隐喻计算知识库的广度和深度不足问题,提出了引入了网络百科资源作为隐喻识别的世界知识库的方法.利用信息检索技术,从概念对应的百科页面中获取概念的背景世界知识,计算两个概念的世界知识的重合程度,作为判断它们是否属于同一个概念域的依据,进行隐喻的识别.实验结果表明在使用百度百科作为世界知识库时,隐喻/常规表达的识别正确率达到81.06%,算法的有效性得到证明.
Aim at the low coverage of current hand-coded knowledge base used in the metaphor recognition, a Chinese nominal metaphor recognition method based on exploiting the online encyclopedia resource is proposed. A noun is represented as a page in the encyclopedia. Capture the world knowledge from the page of encyclopedia, calculates the relatedness between the pages of the two nouns in the nominal reference by applying information retrieval methods. The relatedness determines the reference is metaphorical or not. The experimental results shows that the accuracy of metaphorical/normal references recognition can reach of this method. 81.06% by using BaiduBaike. It supports the validity, efficiency
出处
《计算机系统应用》
2013年第10期8-13,59,共7页
Computer Systems & Applications
基金
国家自然科学基金(61075058)
关键词
名词性隐喻
隐喻识别
百科资源
词语相关度
隐喻可接受度
nominal metaphor
metaphor recognition
encyclopedia resource
words relatedness
acceptability of metaphor