Modified CodeSearchNet (MCSN) Dataset

This dataset is a modification of the CodeSearchNet dataset from CodeXGLUE benchmark, designed for evaluating code summarization models beyond the function level. It explores the impact of function and repository contexts on summary quality. The dataset includes modifications for evaluating at both function and repository levels.

Paper: Code Summarization Beyond Function Level

Dataset Structure:

The dataset contains samples with the following fields:

• repo_name: The repository name.
• method_name: The method name (including class name).
• method_code: The code of a method within the repository (without docstring).
• method_summary: The summary of the method.
• original_method_code: The code of a method within the repository (including docstring).
• method_path: The path to the file with a method within the repository.

Dataset Splits:

• test: 846 samples (806 + 40 samples for few-shot learning)

This dataset is part of a broader study investigating code summarization at various levels (function, class, and repository). The complete study details, code, and evaluation results are available on GitHub: https://github.com/kilimanj4r0/code-summarization-beyond-function-level.