基于事件句关联的新闻主题模型构建方法

摘要

The relevance relationship of news event sentences plays a very important guiding function role in the processing of news topic constraction.A news topic model based on event sentence relevance is proposed.The relevance relationship of event sentences obtained by using the event element of news document will be used as the supervision information in the LDA,then one can get more accuracy probability distribution of news document in the topic space.The experimental results show that ELDA can make better result than LDA in news documents.%考虑新闻事件句关联信息对新闻主题构建的重要指导作用,提出了基于事件分析的新闻主题模型构建方法.利用新闻文本中事件句中包含的事件元素,判断事件句之间的关联,并将这一关联信息作为LDA建模过程中主题分配的指导信息.从而获得不同新闻文档在主题空间上更加准确的概率分布.实验结果表明,提出的新闻主题模型(E-LDA)比LDA主题模型对新闻文本有更好的表达效果.