Abstract: The polarity analysis on objective text is currently a difficult study in the area of natural language processing.Compared with different characters between stock analysts and general network chapters，this paper combines the text structure-based methods of polarity analysis and the model-based methods of polarity analysis，and constructs a combined method of polarity analysis of stock analysts.Results show that the combined method of polarity analysis of stock analysts can improve recall and precision in the consolidated level，respectively reach 90.2%，84.8%.

Key words: polarity analysis, stocks comments, chapter structure, mode

摘要： 对主观性文本的观点倾向性分析是当前自然语言处理领域的一个研究难点，针对股评文章与一般性网络评论相比的不同特点，将基于篇章结构的股评观点倾向性方法和基于模式的股评倾向性分析方法相结合，构建了一种混合的股评观点倾向性分析方法。实验结果表明，混合的股评观点倾向性分析方法可以使查全率和查准率在综合水平上得到显著提高，分别达到90.2%，84.8%。

关键词: 倾向性分析, 股评, 篇章结构, 模式