摘要:情感分析旨在判断文本的情感倾向，而评论质量检测旨在判断评论的质量.情感分析和评论质量检测是情感分析中两个关键的任务，这两个任务受多种因素的影响而密切相关，同一个产品的情感倾向具有相似的情感极性；同时，同一个用户发表的评论质量也具有一定的相似性.因此，为了更好地研究情感分类和评论质量检测任务的相关性以及用户信息和产品信息分别对情感分类和评论质量检测的影响，提出了一个情感分析和评论质量检测联合模型.首先，使用深度学习方法学习评论的文本信息作为联系两个任务的基础；然后，将用户评论及产品评论作为用户的表示和产品的表示；在此基础上，采用用户注意力机制对用户的表示进行编码，采用产品注意力机制对产品的表示进行编码；最后，将用户表示和产品表示结合起来进行情感分析和评论质量检测.通过在Yelp2013和Yelp2015数据集上的实验结果表明，该模型与现有的神经网络模型相比，能够有效地提高情感分析和在线评论质量检测的性能.



Abstract:Sentiment analysis aims to judge the emotional tendency of the text, while the review quality prediction aims at judging the quality of the review. Sentiment analysis and review quality detection are two key tasks in sentiment analysis, these two tasks are closely related by many factors, the reviews on the same product have similar opinion polarity, and the quality of reviews from the same user tend to be similar. Therefore, this study proposes a joint neural model to learn sentiment analysis and quality prediction in order to better study the correlation between sentiment classification and review quality prediction tasks and the impact of user information and product information on sentiment classification and review quality prediction respectively. First of all, this study employs a deep representation learning approach to learn textual information of reviews, serving as the basis to connect two tasks, and then uses the user reviews and product reviews as the representation of the user and the representation of the product, on the basis, a user attention is adopted to encode user information in user representation, and a product attention is used to encode product information in product representation, and finally both user and product representations are jointly integrated for sentiment analysis and quality prediction with attention mechanism. The experimental results on the Yelp2013 and Yelp2015 datasets show that the proposed model can effectively improve the performance of sentiment analysis and online review quality prediction compared with the existing neural network models.



PDF全文下载地址:

http://jos.org.cn/jos/article/pdf/5895