基于细粒度情感的文本挖掘及可视化分析

Text Mining and Visualization Analysis Based on Fine-Grained Sentiment

摘要: 针对当前文本挖掘情感分析缺乏系统性分析研究，建立综合评价指标，系统的分析产品间优劣势，明确改进方向。构建基于细粒度情感分析的模型，首先通过对评论文本进行预处理及分词，再运用LDA主题模型构建属性词典，运用知网情感词词库结合网络新词构建情感词典；接着建立评论有用性规则与情感打分规则，对有用短语打分，获取情感数据集；最后建立四大评价指标，对三款手机进行综合评价及可视化分析。模型数据结果表明，四大指标能够显著突出产品间优劣势，可以帮助生产者更快更准确的了解重点发展方向，也可以帮助消费者更便利的选择钟爱的产品。

Abstract: For the current text mining sentiment analysis, there is a lack of systematic analysis and research, this paper establishes comprehensive evaluation indicators, systematically analyzes the advantages and disadvantages of products, clarifies the direction of improvement and constructs a model based on fine-grained sentiment analysis. Firstly, the comment text is preprocessed and segmented, and then the LDA topic model is used to build the attribute dictionary, and the HowNet sentiment vocabulary is used to build the sentiment dictionary with new words on the Internet; then, in order to obtain the sentiment data set, use the comment usefulness rule and sentiment score rules, scoring useful phrases; finally, four major evaluation indicators are established to conduct comprehensive evaluation and visual analysis of three mobile phones. The results of the model data show that the four indicators can significantly highlight the advantages and disadvantages of the products, which can help the producers to understand the key development direction more quickly and accurately, and also help consumers to choose the products they like more conveniently.