BiLSTM-ATT Chinese Sentiment Classification Model Based on Pre-training Word Vectors

Abstract

Chinese text sentiment classification is an important branch of natural language processing, and deep learning-based sentiment classification has been a hot research topic for sentiment classification tasks in recent years. In this paper, the model introduces pre-training word vectors in the word embedding layer, extracts textual contextual features in the feature extraction layer using a bidirectional long and short-term memory neural network, and adds an attention mechanism to the network structure. The classification layer uses a support vector machine to classify the incoming textual features to achieve the analysis of Chinese text sentiment. The model achieves an accuracy of 87% in a publicly available takeaway review binary classification task.