Berom0227/Detecting-Semantic-Concerns-in-Tangled-Code-Changes-Using-SLMs

Untangling Multi-Concern Commits with Small Language Models

This dataset contains commit data for training and evaluating models on software engineering tasks, specifically focusing on identifying and separating concerns in multi-concern commits.

Dataset Description

This dataset consists of two main configurations:

1. Sampled Dataset (sampled)

• File: data/sampled_ccs_dataset.csv
• Description: Individual atomic commits with single concerns
• Features:
  • annotated_type: The type of concern/change in the commit
  • masked_commit_message: Commit message with sensitive information masked
  • git_diff: The actual code changes in diff format
  • sha: Git commit SHA hash

2. Tangled Dataset (tangled)

• File: data/tangled_ccs_dataset.csv
• Description: Multi-concern commits that combine multiple atomic commits
• Features:
  • description: Combined description of all concerns
  • diff: Combined diff of all changes
  • concern_count: Number of individual concerns combined
  • shas: JSON string containing array of original commit SHAs
  • types: JSON string containing array of concern types

Dataset Statistics

• Sampled Dataset: ~1.3MB, individual atomic commits
• Tangled Dataset: ~7.1MB, artificially combined multi-concern commits

Use Cases

1. Commit Message Generation: Generate appropriate commit messages for code changes
2. Concern Classification: Classify the type of concern addressed in a commit
3. Commit Decomposition: Break down multi-concern commits into individual concerns
4. Code Change Analysis: Understand the relationship between code changes and their descriptions

Data Collection and Processing

The dataset was created by:

1. Collecting atomic commits from software repositories
2. Sampling and filtering commits based on quality criteria
3. Artificially combining atomic commits to create tangled multi-concern examples
4. Masking sensitive information while preserving semantic content

Citation

If you use this dataset in your research, please cite:

@dataset{css_commits_dataset,
  title={Untangling Multi-Concern Commits with Small Language Models},
  author={Your Name},
  year={2024},
  url={https://huggingface.co/datasets/Untangling-Multi-Concern-Commits-with-Small-Language-Models}
}

Scripts and Tools

This dataset includes several Python scripts for data processing and analysis:

• sample_ccs_dataset.py: Script for sampling and filtering commits
• generate_tangled.py: Script for creating tangled multi-concern commits
• clean_ccs_dataset.py: Data cleaning and preprocessing utilities
• show_sampled_diffs.py: Visualization of sampled commit diffs
• show_tokens_distribution.py: Analysis of token distribution in the dataset

License

This dataset is released under the MIT License. See the LICENSE file for details.

Dataset Loading

You can load this dataset using the Hugging Face datasets library:

from datasets import load_dataset

# Load the sampled dataset
sampled_data = load_dataset("Untangling-Multi-Concern-Commits-with-Small-Language-Models", "sampled")

# Load the tangled dataset
tangled_data = load_dataset("Untangling-Multi-Concern-Commits-with-Small-Language-Models", "tangled")