An Efficient Method for Text Classification Task

Abstract

Text classification has always been a research hot-spot in the field of natural language processing. For the neural network input matrix, only the word vector of the word level is extracted, which ignores the expression of the overall semantic features of the text level, resulting in insufficient representation of text features and affecting accurate classification. In this paper, a text representation matrix combining word2vec and LDA topic models is proposed. Combining word meaning and semantic features, inputting LSTM for text classification, and introducing Attention mechanism to improve LSTM model, LSTM-Attention model is designed. The experimental results show that the LSTM classification model has better classification result than the traditional machine learning model, and the LSTM model with the Attention mechanism has a certain degree of improvement compared with the classical text classification models.