BATAE-GRU: Attention-based Aspect Sentiment Analysis Model

Abstract

Aspect-level sentiment classification aims to identify the sentiment polarity of a given aspect. However, most of the past methods do not analyze the role of words well, and the connection between context and a given aspects is not well realized, which greatly limits the effectiveness of the model. In this paper, we have designed a model based on the attention mechanism. First, the word embedding and aspect embedding are represented by pre-trained BERT coding. Next, we use the recurrent neural network to obtain the hidden state. Then, the context and aspect are related through the attention mechanism. Finally, the experiments were conducted on 3 data sets widely used in the field of sentiment analysis. The BATAE-GRU model was compared with several current advanced models. The results showed that the experimental results of the BATAE-GRU model were better than other models; Compared with the ATAE-LSTM model, the accuracy of the model in two comparative experiments has been improved by 6.9% and 9.9% respectively.