Aspect-Based Attention LSTM for Aspect-Level Sentiment Analysis

Abstract

With the rapid development of the mobile Internet, how to efficiently extract information from the massive data in cyberspace and use it has become a key concern of all walks of life. Sentiment analysis aims to dig out the subjective emotional information contained in large-scale texts. It has become a hot research topic in the field of natural language processing and has important application value in many fields. This paper focuses on the aspect-level sentiment analysis problem, which is an important sub-problem in fine-grained sentiment analysis. The purpose of aspect-level sentiment classification is to dig out the sentiment polarity of users' opinions on a specific aspect in the text. This paper proposes a sentiment classification method combining BiLSTM and aspect attention module. BiLSTM has fewer parameters, faster model training, and can effectively extract deep-level information of text; combining the attention mechanism with aspect information can fully extract specific aspects of information. The experimental results show that the F1 score that solves the problem of aspect term extraction and aspect sentiment classification on two data sets at the same time, compared with the existing sentiment analysis method, obtains a better classification effect.