摘要
针对电子商务中的商品评论信息过载问题,运用情感计算理论,通过挖掘商品评论信息中的商品特征及相应的情感褒贬态度,为消费者提供一个商品特征粒度上的情感分析结果,从而帮助消费者从庞杂的商品评论中快速获取有效信息。系统首先采集指定商品的评论集并挖掘商品特征,然后结合情感语料库和词汇相似度计算,利用依存关系找到特征-极性词对以及程度副词和否定词。基于以上结果,考虑程度副词的强度,以及程度副词和否定词共现语序不同造成的语义差异,提出了商品评论情感倾向程度的计算方式。最后,进行系统实现并验证算法的有效性。实验结果表明,系统具有良好的应用效果。
To solve the product review information overload problem in e-commerce,affective computing theory is applied to provide a sen-timent analysis result for consumers in product feature granularity by mining the product features from product reviews information and the cor-responding praise and derogatory sentiment attitude so that to help the consumers get useful information rapidly from large amount of product reviews.The system first collects the assigned product reviews set and mines the product features,then performs calculation in conjunction with emotional corpus and word similarity,and finds out the feature-polarity word pair as well as degree adverb and negatives according to depend-ence relation.Based on the above results,considering the strength of degree adverb and the semantic differences caused by different word order of co-occurring of degree adverb and negatives,we present the calculation means of the sentiment inclination degree of product reviews.Final-ly,the system is implemented and the effectiveness of the algorithm is verified.Experimental result demonstrates that the system has good ap-plied effect.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第12期39-44,共6页
Computer Applications and Software
关键词
电子商务
商品评论
关联规则
句法分析
情感计算
E-commerce
Product reviews
Associate rule
Syntax analysis
Affective computing
