中文摘要：

针对现有的关于网络舆情内容进行情感分析的研究不能满足舆情情感深度挖掘的需求，提出一种基于概率潜在语义分析（ PLSA）的网络舆情话题情感分析方法，利用PLSA模型对不同时间段上的网络舆情话题进行子话题提取和情感词表构建，综合考虑修饰词对情感词的影响以及情感词对子话题的贡献程度，最终得到一个时间序列上各个子话题的情感倾向值以及整个话题的情感变化趋势。实验结果证明该方法不仅可以描述同一个子话题随时间的情感演化过程，还可以描述话题情感随子话题维度和内容的演变情况。

英文摘要：

As the existing sentiment analysis of public opinion on Internet cannot meet the demand for deep excavation of public opinion sentiment, we propose a topic sentiment analysis method based on PLSA. By this method, we extract the sub-topic of public opinion in different period with PLSA and build sentiment vocabularies. Considering the impact of modifiers and the contribution of the sentiment words to the sub-topic, we obtain every sub-topic sentiment orientation in different period as well as the whole topic sentiment evolution. Experiments show that this method cannot only describe the sentiment evolution of one sub-topic over time, but also describe the whole topic sentiment evolution with the change of sub-topic dimensions and content.