lbhsos / graduation-project

Development of An Automatic Classification System for Game Reviews Based on Word Embedding and Vector Similarity

For refactoring

1. Refactored locally and then tested.

2. Tested file might need to merge by hand.

3. Test the merged file on each branch locally, and merge it to the master.

graduationproject

We thought that game software needs a quick feedback to maintain the software. You can see the most interest of users, through game review category classification.

For data collection

we crawled reviews from the Google Play Store to collect data

we used Selenium

fileName: [local]-crawl.py

Package

We used gensim for word2vec and konlpy for data preprocessing

Classificate game reviews using word2vec

We categorized game reviews as payment, account, configuration, server, system, directing, character, etc.

Each category contains about nine sub-words that can represent categories.

To find the categories, we internally weighted the matrix and the TDM document

fileName: [local]-refactoringCategory.py

Satisfaction measurement using CNN

We classified as satisfaction, normal, and dissatisfaction.

fileName: [local]-textcnn.py, train.py, test.py, data.py

Server setting

We used Amazon Web Services to build the server and Flask web framework.