Aspect-Based Sentiment Analysis on Mobile Game Reviews Using Deep Learning
Jiaxin Song · Institutional Repositories DataBase (IRDB) · 2020
This paper proposes an aspect-based sentiment analysis method on mobile game reviews using deep learning, which can make better use of massive mobile game reviews data to judge users' emotional tendencies for different attributes of the game at a finegrained level.Specifically, there are three models in our sentiment analysis method.The baseline model includes Bi-LSTM, FCN, and CRF for sentiment collocation extraction, matching, and classification.The iterative model updates the neural network structure and effectively improves the model's recall rate in the experiments.The joint model is based on the information passing mechanism and further improves the comprehensive performance of the model.We crawled more than 100,000 game review items from two wellknown Chinese game review websites Bilibili and Taptap and manually annotated 3,000 items to construct the experiment dataset.Several experiments have been carried out to evaluate our methods.The experimental results show that our methods have achieved good results.