Abstract: This paper presents a model for analyzing polarity of texts based on semantic segmentation. In accordance with the characteristic of text structure, the text is divided into several semantic paragraphs based on semantic similarity. Conditional random fields model is used to extract the contents of subjective and identify polarity of sentiment in every semantic paragraph, and the weight of each semantic paragraph is calculated. So the sentiment of text is determined by weights and values of polarity of semantic paragraph. Experiments show that compared with traditional machine learning methods, this method can improve the accuracy of text sentiment analysis.

Key words: semantic segmentation, sentiment analysis, conditional random model

摘要： 提出了一种基于语义的观点倾向分析方法。按照文本结构特点，依据语义相近的原则，将文本分割为若干语义段，对语义段采用条件随机场模型进行主观内容提取和观点倾向识别，计算各个语义段的权值，确定文本的观点倾向。实验表明，与传统机器学习方法相比，该方法能有效提高文本观点倾向分析的准确率。

关键词: 语义分割, 情感分析, 条件随机场模型