中文摘要：

[目的/意义]通过分析新闻评论的舆情演化,有助于跟踪民众对社会事件的态度看法。但当前的新闻舆情演化分析方法中存在着不能精确识别新闻焦点及焦点下的评论等问题。[方法/过程]在焦点情感模型（SSCM）的基础上引入新闻报道的时间信息,提出了一种基于焦点情感混合模型的新闻情感演化识别方法（TSSCM）。该方法具备同时建模新闻焦点、时间和评论情感的能力,能够产生文档-焦点、焦点-特征词、焦点-时间和（焦点,时间）-评论情感词4个分布。通过词汇间的点间互信息计算（焦点,时间）-评论词分布内每个焦点在不同时间段的情感值,从而高效地表达出不同新闻焦点的评论情感演化过程。[结果/结论]在5个新闻数据集上分别实验验证了TSSCM模型,并与当前流行的方法进行了对比,实验结果表明TSSCM模型在新闻舆情演化识别方面达到了良好的效果。

英文摘要：

[ Purpose/Significance ] By analyzing the evolution of public opinion of news comment, it helps to track people＇s attitude to- wards social events. In current research, there is a problem that the news subtopics and their comments can not be accurately extracted. [ Method/Process] In this paper, the time information of news report is introduced into the Subtopic Sentiment Combining Model （SS- CM ）, and a novel Time-based Subtopic and Sentiment Combining Model （TSSCM） is proposed. TSSCM has the ability to jointly model news subtopic, time and comment sentiment, and it is capable of generating 4 distributions of document-subtopic, subtopic-word, subto- pic-time and （subtopic, time） -comment sentiment word. By computing the comment sentiment of every news subtopic on different time segment using the method of Point-wise Mutual Information, the comment emotion evolution process of different news subtopic can be ex- pressed efficiently. [ Result/Conclusion] The model is tested on 5 news data sets. The experimental results demonstrate that TSSCM out- performs the state-of-the-art approaches.