Research on Semantic Sentiment Analysis Based on BiLSTM

Abstract

This paper proposes an emotion classification model based on attention mechanism and Bi-LSTM. Firstly, the texts were trained into word vectors by word2vec model. Secondly, Bi-LSTM network was used to learn the semantic information of context. Then attention mechanism was added to extract the relatively important part of the texts for emotion classification. Finally, the activation function was used to classify the texts. This paper made an experiment of emotion classification based on IMDB data sets. The experimental results show that the proposed model is better than the traditional machine learning method and recurrent neural network in accuracy, recall and F 1 value. This proves that the AT-BiLSTM model proposed in this paper can effectively improve the effect of emotion classification, and has certain practicability.