Abstract: With the rapid development of the Internet, more and more people like to comment on goods in micro-blog and forum, so there are a large number of reviews on the network. In order to mine the aspect and sentiment in comments simultaneously for consumers and manufacturers, this paper presents a novel labeled joint aspect sentiment model (Labeled-JAS) for movie review data. Labeled-JAS model can mine aspects and sentiments from movie review data, and the underlying assumptions of Labeled-JAS model are that sentiment distribution depends on aspect distribution and words are the smallest units of sampling. This paper also combines Labeled-JAS model with dictionary-based method, the experimental results show that this method performs well on the movie review data set of COAE2016 task 2.

Key words: COAE2016, dictionary, labeled joint aspect sentiment model, movie reviews

摘要： 随着互联网的蓬勃发展，越来越多的人喜欢在微博和论坛上对商品进行评论，致使网络上存在着大量评论数据。为了同时挖掘评论数据中所谈论的方面以及评论者对这个方面的观点，用于指导消费者的消费和生产厂家对商品的改进，面向电影评论数据提出了一个标签方面情感联合模型。该模型可以同时挖掘出电影评论数据中所评论的方面以及对这个方面的情感，并且假设情感分布依赖于方面分布，词是采样的最小单位。通过将传统基于词典的方法和模型联合使用，在COAE2016任务2的电影评论数据集上进行测试，实验结果表明，此方法取得了较好的结果。

关键词: COAE2016, 词典, 标签方面情感联合模型, 电影评论