.png){kind=icon}
.png){kind=icon}
Return
HypeAssign: Hypergraph contrastive learning for issue assignment
.png){kind=icon}
PRE
AI
DOI：10.1007/s10664-026-10877-5.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Issue assignment is a critical task in software maintenance, aiming to recommend appropriate developers for newly reported issues automatically. Existing studies have introduced graph-based models to capture the structural relationships between issues and developers. However, on the one hand, they rely solely on sparse issue–developer interactions and overlook the potentially rich interactions in software development activities. On the other hand, traditional pairwise graph modeling focuses on local neighborhood information while overlooking high-order relational structures that may carry deeper collaborative semantics. To address these limitations, we propose HypeAssign, a novel hypergraph contrastive learning framework for issue assignment. Specifically, we extract multi-relationships among issues, developers, and source code files to construct a unified hypergraph, aiming to enrich the modeling of issue–developer interactions. Then, we introduce a cross-view contrastive learning strategy that aligns local neighborhood information and global high-order relational structures. Finally, to evaluate the performance of HypeAssign, we conduct extensive experiments on five open-source datasets. The results show that HypeAssign achieves an average improvement of 50.27% in top-1 hit rate and 39.49% in MRR over the best-performing baseline across these datasets.
Keywords:
Issue Assignment
Multi-Relationship
Hypergraph
Journal
IF:
3.6
Papers:
2.0K
Citations:
5.3K