CAJ | 학술논문

针对微博中的海量产品评论信息，提出了一种基于模糊观点词的产品评论情感极性和强度计算方法。该算法运用规范化的TFIDF加权方法提取产品特征，基于知网构建模糊观点词词库，应用BMI （Balanced Mutual Information）方法进行特征词和观点词关联度计算，因而有效解决了微博产品评论中特征-观点对的提取问题。通过微博文本影响力分析，结合对微博文本中的情感语义因素定量计算，提高了微博产品评论情感分析的准确率。给出了应用该方法的具体步骤，通过实验分析发现本文构建的算法在各方面的表现都处于不错的水平并具有很好的应用性。
침대미박중적해양산품평론신식，제출료일충기우모호관점사적산품평론정감겁성화강도계산방법。해산법운용규범화적TFIDF가권방법제취산품특정，기우지망구건모호관점사사고，응용BMI （Balanced Mutual Information）방법진행특정사화관점사관련도계산，인이유효해결료미박산품평론중특정-관점대적제취문제。통과미박문본영향력분석，결합대미박문본중적정감어의인소정량계산，제고료미박산품평론정감분석적준학솔。급출료응용해방법적구체보취，통과실험분석발현본문구건적산법재각방면적표현도처우불착적수평병구유흔호적응용성。
For mass product reviews in Chinese microblogging, the paper builds a microblogging product review mining method based on fuzzy evaluation words ontology. Normalized TFIDF weighting scheme is applied to extract the most informative noun patterns to represent the product features, then we use the established fuzzy evaluation words ontology and BMI method to extract evaluation words for product features, and build the influence calculation method of microblogging text, construct the sentiment computing method of product reviews in microblogging, and finally we find that our system shows remarkable performance improvement over the baseline menthod through experiments and data.