Deep Bug Triage Model Based on Multi-head Self-attention Mechanism

Abstract

AbstractIn the course of software maintenance and development, bugs are inevitable. At present, Bug tracking system uses bug reports to match bug with fixers. However, the previous bug triage model relies too much about the quality of the text of the bug report, introduces a lot of redundant information in natural language, and ignores the fixer community factor where the meta-field of the bug report, which makes the model performance worse. Aiming at the above problems, we propose a multi-head self-attention deep bug triage (MSDBT), which considers the text content of the bug report and generates a sequence of fixers with the same meta-field. The features of the input text and the fixer sequence are extracted by Bi-directional Long Short-Term Memory network. The multi-head self-attention mechanism is used to perform parallel attention calculation among the internal input elements. The model weakens the redundant information in the bug report, and further quantifies the influence of fixers with similar activities on bug triage through fixer sequence. We conducted texts on four open source software projects. We can get the MSDBT has clear strength over the previous model in recall index.KeywordsBug tracking systemBug reportBug triageDeep learningFixer communityMulti-head self-attention