Web文档评价对象抽取研究

Research on Evaluation Object Extraction from Web Document

针对用户评论信息,基于句子上下文相关线索和上下文无关线索信息,应用句法分析、点互信息方法,构造领域词典及主张词典,对产品评论中的评价对象进行抽取。实验结果说明,与取距离评价词最近的词作为评价对象的Baseline方法相比,该方法能提高准确率和F度量值。

This paper considers both the contextual clues and context-independent statistical clues,using syntactic analysis and method of point of mutual information,building related dictionaries to extract the evaluation object in products comments.Experimental results show that this method is better than Baseline method which gets the latest word as the evaluation object in precision rate and F-measure.