Aspect Level Sentiment Classification Based on Viewpoint Information Unit

Aspect level sentiment classification is the process of discriminating the sentiment orientation of different aspects expressed in the text. Due to the non-standardization of Chinese comments, there are different levels of noise in the text, which affects the classification results of aspect level sentiment analysis. In this paper, an Aspect level sentiment classification method based on evaluation unit is proposed. Firstly, based on the results of part of speech analysis, dependency syntax analysis and semantic dependency analysis are used to obtain the relationships between words, so as to make relevant rules for viewpoint information recognition. Then, the redundant text is deleted by filtering the hypothetical and declarative statements of the key sentences, and the method of constructing the evaluation collocation table combined with the structure diagram of the comment object is completed to identify the implicit evaluation units. Finally, aspect - level sentiment classification is achieved based on sentiment dictionary.