Software Bug Report Triaging Based On Self-Attention Mechanism

As the architecture of software systems becomes more complex,the number of Bug reports in software projects is increasing,which greatly increases the labor and time costs of software projects.Therefore,since 2004,more and more automatic or semi-automatic Bug report assignment algorithms have been proposed to replace the original human triager.Although research in the field of Bug report assignment has made great progress,there are still many deficiencies.First of all,most of the relevant research only focuses on the accuracy of the algorithm,and there is no comprehensive synthesis to make a trade-off between the accuracy and the time consumption of the algorithm.But in the industrial environment,it is undoubtedly more effective that complete the assignment of Bug reports better in a shorter time.Second,many studies do not make full use of the information of other metadata features in the Bug report,and they only rely on the text description of the Bug report for Bug assignment,which has caused a great waste of resources.In response to these two problems,this paper conducted the following research:(1)This paper proposes a Bug report assignment algorithm based on the self-attention mechanism.The algorithm uses the self-attention block to replace the traditional recurrent neural network and convolutional neural network to extract text features.The experiments on the five datasets of Mozilla,Netbeans,GCC,Eclipse and OpenOffice show that the algorithm proposed in this paper can achieve higher accuracy than other Bug report assignment algorithms in a shorter time;(2)This paper proposes a Bug report assignment algorithm based on multiple features.The product and component features of the Bug report are combined into the algorithm through a meta-feature network,which further improves the advantages of the algorithm in this paper compared to other Bug report assignment algorithms.At the same time,we also propose a controllable cost sensitive layer to further improve the performance of the algorithm in data imbalance.