摘要： 受语言固有的模糊性、随机性以及传统文本特征词权重值计算方法不适用于情感词等因素的影响，文本情感分类的正确率很难达到传统文本主题分类的水平。为此，提出一种基于情感词属性和云模型的情感分类方法。结合情感词属性和简单句法结构以确定情感词的权重值，并利用云模型对情感词进行定性定量表示的转换。实验结果表明，该方法对情感词权重值计算是有效的，召回率最高达到78.8%，且与基于词典的方法相比，其文本情感分类结果更精确，正确率最高达到68.4%，增加了约9%的精度。
关键词:
观点挖掘,
文本挖掘,
情感分类,
云模型,
情感词属性,
文本特征提取
Abstract: In the era of big data, how to obtain valid information from the Web becomes a keen topic for business, government, and research workers. User’s opinion mining becomes a research topic for the area of Natural Language Processing(NLP) and text mining. However, due to the inherent fuzziness and randomness of language, as well as the traditional term weight value calculation method is not suitable for the sentiment word and other factors, the text sentiment classification accuracy is difficult to achieve the performance of traditional text subject classification. To solve these problems, this paper proposes a sentiment classification method based on sentiment word attributes and cloud model. It calculates weight of sentiment words by combining attributes and syntactic structure of sentiment words, and converts qualitative and quantitative of sentiment words based on cloud model. Experimental results show that this method to calculate weights of sentiment words is valid, and the recall rate is up to 78.8%. Text sentiment classification results are more accurate than that based on dictionary, the correction rate is up to 68.4%, and the accuracy is increased by about 9%.
Key words:
opinion mining,
text mining,
sentiment classification,
cloud model,
sentiment word attributes,
text feature extraction
中图分类号: