基于概率图模型的文本对象情感分析

摘要

情感分析旨在从文本数据中自动识别主观情感，即文本中表达的观点、态度、感受等，在线评论通常都涉及特定的对象，通过在JST模型基础上加入对象层提出了一种无监督的对象情感联合模型（UOSU model），UOSU模型对每个词同时采样对象、情感和主题标签，最终得到各个主题的对象情感词以及文本的对象情感分布。在汽车评论数据集上进行的情感分类实验取得了74.19%的精确率和73.97%的召回率。%Sentiment analysis aims to automatically detect the subjective sentiment such as opinions, attitudes and feelings in texts. Online review usually relates to the specific object, in order to detect object information from texts, this paper proposes an unsupervised object and sentiment unification model (UOSU model), which has added a plate for object based on the JST model. UOSU model is fully unsupervised and detects object, sentiment and topic simultaneously from text. Using the UOSU model can achieve words generated by the specific object, sentiment and topic. Besides, the object and sentiment for a text can be obtained at the same time. The model is evaluated on the car review dataset to classify the review sentiment polarity, and the sentiment classification experiments achieves accuracy of 74.19%and recall rate of 73.97%.