This repository makes the Github Sentiment Dataset compatible with Kaiaulu for sentiment analysis of developer communication across open source software projects.

The sentiment-labeled data comes from Novielli et al. (2020), "Can We Use SE-specific Sentiment Analysis Tools in a Cross-Platform Setting?", and covers 7,122 GitHub comments manually labeled positive, negative, or neutral.

Download the Github 2020 CSV. The dataset contains github_gold.csv, which has 7,122 GitHub comments with three columns: ID, Polarity, and Text

The CSV alone has no author, timestamp, or project context. To recover that context, the Github 2020 CSV is joined to the GHTorrent MSR 2014 dump via comment_id. This yields main project repos with sentiment-labeled comments that can be downloaded and analyzed through Kaiaulu.

python -m venv .venv
source .venv/bin/activate

pip install pandas sqlalchemy mysql-connector-python pymysql pyyaml

Update the MySQL connection variables at the top of each notebook to match your local setup:

MYSQL_HOST     = "localhost"
MYSQL_PORT     = 3306
MYSQL_USER     = "root"
MYSQL_PASSWORD = "ADD_YOUR_PASSWORD_HERE"
MYSQL_DB       = "github"

Clone Kaiaulu locally. Set KAIAULU_REPO in Notebooks 3 and 4 to your local Kaiaulu path before running.

A GitHub personal access token is required by Kaiaulu's download vignettes. To create one:

1. Go to GitHub → Settings → Developer settings → Personal access tokens → Tokens (classic)
2. Enable the public_repo scope
3. Save the token to ~/.ssh/github_token

• Novielli et al. (2020). "Can We Use SE-specific Sentiment Analysis Tools in a Cross-Platform Setting?"