Context-Aware Deep Learning for Effective Unit Test Case Generation and Repair.

This thesis addresses the challenge of creating effective automated software unit test cases. Current methods are limited as they rely only on the program’s source code. This research develops new techniques that use additional context such as developer written descriptions, code comments, and assertions to generate more accurate test cases. It also introduces a method to automatically repair outdated tests when the software changes. This research helps software developers save time and improve the quality of their code through test case generation.