Understanding Mobile Game Reviews Through Sentiment Analysis: A Case Study of PUBGm

This project designs a comprehensive framework that employs topic modeling and sentiment analysis to extract insightful keywords from a vast collection of reviews. The experiment was conducted on a dataset containing six million English reviews collected up to March 2023 for the mobile game PUBGm from Google Play.

Requirements for running the experiment

• python 3.7.3
• pytorch 1.2.0
• transformers 4.1.1
• numpy 1.16.4
• tensorboardX 1.9
• tqdm 4.32.1

The codes are derived from: Yu, Y., Dinh, D. T., Nguyen, B. H., Yu, F., & Huynh, V. N. (2023). Mining Insights from Esports Game Reviews with an Aspect-Based Sentiment Analysis Framework. IEEE Access.

Requirements for LDA

• python 3.7.3

• gensim 4.0.1

• spacy 3.0.6

  python -m spacy download en_core_web_sm

• scipy 1.6.2

Dataset

The collected dataset can be downloaded at: https://drive.google.com/file/d/1uzB131qYmW3dwHNavsRMxQofvdhBD8OB/view?ts=6517ecba

Environment

• OS: Ubuntu 20.04.2 LTS
• GPU: NVIDIA A40
• CUDA: 10.0
• cuDNN: v7.6.1