Towards Quality Assurance of Natural Language in Code

The quality of code in software development is critical to ensuring robust, functional, and maintainable applications. While static analysis techniques can improve code quality, they often overlook the natural language (NL) components of code, such as comments and logging statements, which play an essential role in program development and maintenance. In this thesis, we aim to address quality assurance for these NL elements through two primary objectives. First, we plan to propose a new benchmark to facilitate consistent comparisons among existing methods. Second, we tackle logging statement quality by identifying a broader range of defect types and introducing an automated detection and repair tool. This work addresses critical limitations in current research, which often focuses on single defect types and lacks repair functionality, by enabling a comprehensive and practical approach to NL quality in code. Our contributions provide a foundation for ongoing improvements in code maintainability and contribute to a more reliable software development lifecycle.