摘要
提出了一种只利用Wikipedia的链接结构化信息度量词汇间语义相关性的新方法,在语义相关性的计算过程中,综合考虑了两种指向的共享链接(指入型、指出型)和三种链接相关的类型(直接链接相关、间接链接相关、传递链接相关)。利用多个通用的测试数据集与当前若干主流语义相关性度量方法进行了实验比较,结果表明本文方法在不需要进行任何的文本处理的情况下取得了前所未有的好效果。
A new semantic relatedness measurement technique between words based only on link structure information of Wikipedia was provided. During the process of relatedness computation, the positive effects of two-directional shared links (incoming links, outgoing links ) and three kinds of link-relevance types (direct link-relevance, indirect link- relevance, transitive link-relevance) have been taken into account comprehensively. Using several widely used test sets as benchmark, we compared our measure with several popu!ar methods in computing semantic relatedness. Experiment results showed that our method made an unprecedented excellent result without any text processing.
出处
《情报学报》
CSSCI
北大核心
2013年第4期385-389,共5页
Journal of the China Society for Scientific and Technical Information
基金
浙江省自然科学基金项目(LQ12F020047)