Sentiment Analysis based on GloVe and LSTM-GRU

Abstract

Sentiment analysis technology generates linguistic understanding from the perspective of machines through the processing and analysis of large quantities of data, which is a hot research direction in the field of artificial intelligence. In this work, deep learning is used to build a sentence sentiment analysis system based on word vector space model and Recurrent Neural Network. Specifically, the GloVe word model is adopted here with the aim of making the best use of the global and local information for the training corpus,. In order to overcome the defect that traditional recurrent neural network cannot learn the long-term information of text, a neural network model combining LSTM and GRU is proposed in this work. Through the contrast experiment of RNN, LSTM and GRU, comparing the training time and test accuracy indicators, it is found that LSTM-GRU model performs best.