中文摘要：

由于传统褒贬二值分类模型缺乏对文本主题之间以及主题与观点持有者之间的关系挖掘,不能很好的处理具有不规则、口语化、极性强等特点的评论文本.通过提取网站的文本评论,对评论对象进行结构化处理,以How Net公布的情感词典为基础,完善了评论情感倾向性词典.结合五元组模型量化文本情感信息,建立了适合处理评论文本的模型,深度挖掘了用户对商品或消费行为的主观感受.并通过实验验证了该模型的准确度和有效性.

英文摘要：

Considering that the traditional binary classification model cannot effectively excavate the relationship between view-holders and different themes, and cannot handle reviews-text with irregularity,colloquial, polarity and other characteristics, the study extracts text reviews from website and conducts structural processing on the review objects based on the emotional lexicon from How Net. Thus, the review emotional tendency lexicon is refined. Combined with the quintuple model and characteristics of reviews, we established a model for processing reviews and analyzing user＇s subjective feeling from this consuming behavior. The accuracy and effectiveness of the model are verified by experiments.