Widsom of the Gaming Crowd

Steam Feature Reviews

This project implements a feature-based summarisation algorithm, designed for summarisaing Steam reviews.

Feature Summaries

Ever enjoyed a game that everyone else hates?

One simple explanation is that different individuals value different aspects of a game to different extents. A shooter's lack of story might be a dealbreaker to one player; whereas another might only care about the gameplay.

This creates a challenge when summarising crowd-sourced reviews, such as on Steam. Providing a single number, e.g. "87% reccomended" misses the nuance of preference. To give extra context, we can provide ratings for each aspect. For example:

Shooter Inc.
Total: 74% positive.

• Gameplay - 87% positive.
• Graphics - 61% positive.
• Story - 21% positive.

This project provides a baseline algorithm to generate these summarises from text reviews via topic extraction and sentiment analysis.

This work contributed to The Wisdom of the Gaming Crowd (Robert Jeffrey, Pengze Bian, Fan Ji, and Penny Sweester, 2020), submitted to CHI PLAY 2019. It was originally completed as part of the Individual Research Project course (COMP3770) at the Australian National University.

Project structure:

• report.pdf - project report from original submission.
• download_reviews/download_reviews.ipynb - download reviews from Steam.
• feature_mining/main.ipynb - generate feature summaries from downloaded reviews.
• review_data.zip (LFS) - review data used in original project report. Password is available upon request.

Installation & Set-up

Python

This project uses Python3. Instructions are provided for Anaconda setup below:

Create the environment:

conda env create -f environment.yml

Next, activate the environment:

conda activate steam-feature-reviews

Finally, open jupyter notebook:

jupyter notebook

If installing without anaconda, the following libraries are required:

• jupyter
• pandas
• nltk
• requests
• tqdm

Generating Feature Summaries

First, identify the Steam app_id for the video game(s) of interest. This can be obtained from the URL of the Steam page. For example:

https://store.steampowered.com/app/570/Dota_2/
gives
app_id = 570

Secondly, use download_reviews.ipynb to gather review data. >10k reviews (100 pages) are easily handled and makes a good starting point for analysis. More data can be easily gathered by running the script again due to the restart capabilities.

Finally, use feature_mining/main.ipynb to summarise the reviews.

Contributing

This repo is currently an archive, so new features will not be added. However, forks are welcome!