Learning Developer-Code Contribution Relationship via Heterogeneous Graph Attention Network for Repositories Oriented Code Readability Assessment

project
Jan 2022

To improve the performance of code readability assessment and enhance the practicality of existing assessment approaches, we propose a repository-oriented code readability assessment method based on Heterogeneous Graph Attention Network (HAN) to learn developer-code contribution relationship.

  • Automatically assessing code readability is important for project maintenance and evolution. However, existing code readability research only focuses on single files' assessment. These studies fail to utilize developer-code contribution relationship, which refers to the valuable information underlying code files and their developers in the context of code repository.

    • Paper

    • PyTorch
    • JavaParser
    • Heterogeneous Graph Attention Network
    • HPC