Automatically Extracting Bug Reproducing Steps From Android Bug Reports

Many modern software projects use bug-tracking systems (e.g., Bugzilla, Google Code Issue Tracker) to track software issues and help developers reproduce these issues. There has been recent work on automatically translating the natural language text (i.e., steps to reproduce) of bug reports to reproducing scripts, targeted at Android apps, to facilitate app debugging process. The scripts describe the event sequences leading to the app issues and thus can be reused for testing newer versions of the apps. However, existing techniques require manually providing the text description of steps to reproduce for generating reproducing scripts, which is a non-trivial task because natural language text in bug reports can be complex and contain much information irrelevant for bug reproduction. In this paper, we propose an approach that can automatically extract the text description of steps to reproduce (S2R) from bug reports to advance automated software issue diagnosis and test script reuse. The approach is implemented as a tool, called S2RMiner, which combines HTML parsing, natural language processing, and machine learning techniques. We have evaluated S2RMiner on 1000 original Android bug reports. The results show that S2RMiner can extract S2R with high accuracy.