BEAT: Considering question types for bug question answering via templates

摘要

Software bugs are ubiquitous in the process of software development and evolution. To accelerate bug fixing, developers need to quickly obtain bug information and understand the bugs at hand. Many existing approaches often do not try to understand the semantic information of bug data and just directly use keyword matching techniques to obtain more bug information. However, they often fail to fully express the users’ intents, and the results obtained are often irrelevant to the input queries. To alleviate this problem, this paper presents a novel approach named considering question types for bug question answering via templates (BEAT). We utilize templates to understand the natural language questions through factual triples. These templates can fully express users’ intents and generate the structured queries based on these intents. Then, answers are automatically generated by executing the structured query over the prepared RDF file. Empirical study demonstrates that BEAT is effective to automatically generate the answers, and the F1-score values of Mozilla and Eclipse project are 0.76 and 0.73, respectively, which are better than existing Q&A approaches.