Mining Event Logic Graph from Open Q&A Site for Automated Program Repair

Automated program repair (APR) has been a hot topic due to its possible usage among large group of program-mers. To reduce the patch search space, APR tools usually rely on external sources like StackOverflow. However, most existing approaches simply associate a bug with a single post/patch, which has a negative impact on the performance. This paper proposes the idea of automated program repair event logic graph (APR-ELG), which links relevant bug reasons and possible patches with graph structure. Preliminary studies on Defects4J have shown that APR-ELG can correctly repair several new bugs.