Aspect Level Sentiment Classification Based on Double Attention Mechanism

Abstract

Aspect sentiment classification is a fine-grained sentiment classification method, which is used to identify the sentimental polarity of a given aspect word in one sentence. Among the existing aspect-level sentiment classification methods, the deep learning model with the attention mechanism solves the problem of key word recognition in sentiment analysis and achieves good results. However, the effect of sentiment classification is not good in complex sentence structure and informal expression. In the deep learning model of aspect-level sentiment classification, this paper combines internal attention with external attention, and constructs an aspect-level sentiment classification model based on double attention, which consider the internal structure of the text as well as the external attention concerns of people. Experiments were conducted on SemEval2014 and Twitter datasets. Compared with the classical classification methods, the recognition accuracy of the proposed algorithm in this paper was improved by about 1%, and better results were achieved.