|GitHub Updates Issues Picker|
|Written by Alex Denham|
|Friday, 28 February 2020|
GitHub has updated a tool that identifies areas for work on open source projects that are relatively easy and would be a good place to start contributing. The tool uses a combination of a machine learning model that has been trained to identify easy issues, and an associated list put together by project maintainers.
The possibilities are listed as beginner-friendly issues in the 'contribute' section for projects on GitHub, a facility that was first available last year as recommendations based on labels that were applied to issues by project maintainers. The GitHub team analyzed its data and came up with a list of about 300 label names used by popular open source repositories that described either "good first issues” or “documentation”. This search found suitably labelled issues in around 40 percent of repositories.
The updated version identifies issues in about 70 percent of repositories falling into the category of being suitable for beginners. This greater coverage has been achieved using a machine learning model that automatically infers labels for hundreds of thousands of candidate samples. Discussing the updated version, GitHub's Tiferet Gazit said:
"There is a tradeoff between coverage and accuracy, which is the typical precision and recall tradeoff found in any ML product. To prevent the feed from being swamped with false positive detections, we aim for extremely high precision at the cost of recall. This is necessary because only a tiny minority of all issues are good first issues."
Going forward, the aim is to improve the issue recommendations by iterating on the training data, training pipeline, and classifier models to improve the surfaced issue recommendations. The team is also adding better signals to repository recommendations to help users find and get involved with the best projects related to their interests. They also plan to add a mechanism for maintainers and triagers to approve or remove ML-based recommendations in their repositories.