【摘要】 用户发表的购后评论,隐含着消费者对产品特征的某种偏好。有效挖掘消费者的偏好信息,有利于企业制定差异化的营销策略。首先,采用最大熵模型识别消费者性别。然后,利用CRF模型、频繁特征和评价词搭配从评论文本中抽取"产品特征─评价短语"对,使用情感词典为评价短语赋值。最后,以产品特征评价值为自变量、用户评分为因变量,利用二元Logistics回归分析构建消费者偏好模型。以亚马逊中国网站的两款热销手机的评论为实验,建立了消费者的总体偏好模型和不同性别消费者的偏好模型。模型显示男性消费者和女性消费者的偏好存在差异,给出了相应的营销建议。更多还原

【Abstract】 Comments published by users imply consumers’ preference to product features. It is helpful for companies to develop differentiated marketing strategies by effectively mining consumers’ preference information. Firstly,the maximum entropy model is used to identify the gender of the consumer. Secondly,CRF model,frequent features and evaluation words are used to extract "product feature-appraisal phrase"from the comment texts,and the emotional dictionary is used to assign a value to the appraisal phrase. Finally,binary Logistic regression analysis is used to establish the consumer preference model by taking the evaluation value of product features as the independent variable and the user rating as the dependent variable. This paper,taking the comments on two kinds of popular mobile phones from Amazon website as experimental corpus,establishes the consumers’ overall preference model and the preference model of different gender consumers. The model shows that preferences are different between male consumers and female consumers. The corresponding marketing recommendations are given.更多还原