知网的话题更新与跟踪算法研究

Algorithm Study of Topic Tracking Based on HowNet and Topic Renewal

话题跟踪是一项面向新闻报道信息流进行已知话题跟踪的信息处理技术,本文在现有的向量空间描述文档的基础上,实现了一个基于话题更新的话题跟踪算法,提出了基于知网的近义动词分析的跟踪算法及基于知网的近义动词分析和话题更新的话题跟踪算法.实验表明话题更新和同义词近义词的消除提高了话题跟踪算法的性能.

Topic tracking, as one of natural language processing technologies, is to track known topic from the information stream of news media. This paper implements an algorithm of topic tracking based on topic renewal through describing the vector space model. In addition, it also presents an algorithm of topic tracking based on synonymic verb analysis of Hownet, and an algorithm of topic tracking which is combined synonymic verb analysis with topic renewal. The experiments show that topic renewal and synonymic verb analysis can improve the performance of the topic tracking system.