RHE: Relation and Heterogeneousness Enhanced Issue Participants Recommendation

Open source software (OSS) platform users frequently join issue discussions in various repositories, and establish numerous co-talk (i.e. cross-issue reference and discussion) relationships between issues both within or cross repositories. In this work, we collect and analyze issue discussion data from GitHub to study the unique features of co-talk relationships, and discover that many participants play a versatile role during issue discussions across repositories. Based on the discovery, we enhance existing bug triaging technologies with the Relation and Heterogeneousness Enhance (RHE) method to include potential participants with cross-repository co-talk histories. RHE integrates co-talk relationship embedding and heterogeneous graph embedding for complex OSS communities. We conduct experiments with real-world data collected from GitHub to show the effectiveness and usefulness of RHE. The results suggest that RHE achieves an improved performance comparing to the baseline approaches.