Review-Issue-Patch Aligned Dataset for Android Program Repair

27335 instances connect user reviews, developer issues, and bug-fixing commits from open-source Android applications. This dataset provides weakly aligned, semantically matched artifacts to bridge symptom-level descriptions with technical fixes. It enables research into defect localization and patch generation under weak supervision from user feedback.

Use Cases

• Training review-augmented program repair models based on semantic alignment between symptoms and fixes.
• Studying cross-level defect localization using weakly aligned review-issue pairs.
• Benchmarking automated program repair techniques under weak supervision from user reviews.
• Analyzing the relationship between observable symptoms in app reviews and technical causes in issue reports.

Strengths

• Contains 27335 aligned instances connecting three artifact types: reviews, issues, and patches.
• Provides a novel, semantically aligned structure for a weakly supervised research problem.
• Released under a permissive CC-BY-4.0 license, facilitating reuse and modification.

Limitations

• Column names and exact row count are not specified, limiting precise understanding of the data structure.
• Alignments are established through semantic matching rather than explicit links, introducing potential noise.
• The dataset's temporal coverage and the specific Android applications included are not detailed.

Provenance

Source

Open-source Android applications.

Collection Method

Connections established through semantic alignment combining representation learning and structured symptom matching.

Time Range

null

Freshness

Last updated on 2026-03 21.

Geography

null