Approach to automatic analysis for relevance among software bug reports

The approaches to the relevance analysis of bug reports were studed based on natural language processing technology, but the precision and recall were hard to improve. The paper proposed the new approach based on both of structure information (including patches, exception stack and code fragments). With these information, the relevance of bug reports detected by using the similarity analysis of the structure and text information. It used 1 000 bug reports form Eclipse to test, and the experimental result shows that it can effectively improve precision rate and recall rate to about 90% by adding similarity analysis of the structure information.