基于混合信息增益算法的文本情感分析

Text Emotional Analysis Based on Hybrid Information Gain Algorithm

摘要: 针对传统信息增益特征选择方法存在的选择偏向性的现象以及未考虑特征元素在不同类别间词频的问题，提出了一种混合信息增益的文本情感分析算法。通过引入逆文档频率系数、类间特征词频系数和卡方统计量系数，对文本数据进行特征选择，使得整个文档中词频信息、每个类之间的词频信息以及重要情感色彩的低频词信息得到有效利用。实验结果表明，采用混合信息增益的文本情感分析方法可以有效地提高特征选择的质量，进而提高文本情感分析的准确率，大约2%~5%。

Abstract: Aiming at the problem of selection bias in the traditional information gain feature selection method and the problem of word frequency between different categories without considering the feature frequency of different elements, a text sentiment analysis algorithm with mixed information gain is proposed. By introducing the inverse document frequency coefficient, the inter-class feature word frequency coefficient and the chi-square statistic coefficient, the text data are feature-selected, so that the word frequency information in the entire document, the word frequency information between each class, and the low-frequency word information of important emotional colors are obtained and used efficiently. The experimental results show that the text sentiment analysis method with mixed information gain can effectively improve the quality of feature selection and improve the accuracy of text sentiment analysis, about 2% to 5%.