Bidirectional Long Short-Term Memory for Sentiment Analysis of Chinese Product Reviews

摘要

Sentiment analysis of online product reviews and other user generated contents is a meaningful research subject for its wide range of applications. Traditional feature-based methods suffer from the limited scope of insufficient features caused by too short length of text. Bidirectional Long Short-Term Memory(Bi-LSTM), a neural network that extracts features of text automated, broadly used in data processing and predictions. Our paper is to explore a way for Bi-LSTM to identify the emotional polarity of product reviews. Our experimental results demonstrate that our proposed method, compared with previously reported model, performs better in precision, recall and F -score evaluation on three reviews data sets respectively.