Aspect Based Sentiment Analysis Based on Interactive Attention Neural Network Model

Abstract:
Aspect-based sentiment analysis is one of the hot research directions in natural language processing. The traditional sentiment analysis technology can only judge the polarity of a sentence as a whole, but cannot judge the sentiment tendency of multiple targets in the sentence separately. Therefore, a new aspect-based sentiment analysis research direction has emerged in the field of natural language processing, which can perform sentiment analysis for multiple targets in a sentence. An interactive attention neural network model (Bi-IAN) based on bidirectional long and short-term memory neural network (BiLSTM) can be used to solve the problem of modeling and sentiment analysis of different targets. Through BiLSTM, different targets and important semantic information of the sentence can be extracted, and then the relationship between objects and emotions can be recognized through the interactive attention mechanism, and finally the emotions of different targets can be classified through the non-linear layer. Experiments were carried out on the data set SemEval 2014 task 4 and the Chinese comment data set. The experimental results show that the accuracy and F1-score of the model are better than the existing sentiment analysis model.