摘要
品牌舆情管理涉及文本、语音等自然语言产物的处理,如挖掘文本内涵的情感、观点等并对其量化,才能进一步分析品牌所处的舆论环境。对自然语言中情感的量化即情感判别分析,针对传统的基于词典的情感分析和基于监督模型的情感分析存在的不足,提出了一种新的情感分析系统,并结合朴素贝叶斯分类算法,提高了情感分析的准确率,并增强了量化分析情感强度的能力。经测试,提出的文本情感分析引擎的情感判别准确率高于常见的分析方法,且不具有非常明显的行业特异性。
Brand public opinion management involves text, voice and other natural language processing, such as mining the emotions and views of the text and quantifies it. The quantification of emotion in natural language is the emotion discriminant analysis. Considering the disadvantage in the traditional sentiment analysis that based on emotional dictionary and supervision model based sentiment analysis system, a new sentiment analysis system was proposed, and combined with the Naive Bayesian classification algorithm, the accuracy of sentiment analysis was improved, and the ability of quantitative analysis of emotional strength was enhanced. The sentiment discrimination accuracy of the proposed text sentiment analysis engine is higher than that of the common analysis method, and there is no decent of accuracy in out-of- sample texts from different industries.
出处
《大数据》
2017年第6期55-64,共10页
Big Data Research
关键词
情感分析
监督模型
朴素贝叶斯
自然语言处理
sentiment analysis, supervised model, naive Bayes, natural language processing
