Automated Summarizing of Bug reports to Speed-up Software Development/Maintenance Process

Abstract

Software’s development process can be optimized by using the knowledge about past information about same kind of product or problem. During development process, software’s bug repository can provide a great deal of easiness for development team. It can be a rich source of information for developers and other members of development team. Bug reports can provide a great deal of assistance for developers during the process of development. But due to the large size of bug repositories, it is sometimes difficult to take advantage of these artifacts in the available time. One way of helping developers to provide summaries of these reports and provide relevant details only. Once it’s decided that this is the required report then one can study the details. We analyzed the previous approaches use for this purpose and realized that there is need of improvement in this research. We used an extractive summarization approach using the unsupervised learning method for this purpose and developed a novel framework to get better results than previous a state of the art systems. As text mining technology advances, many substantial approaches have been proposed to generate optimized summaries for bug reports. In this paper, we have proposed an extractive based methodology for the generation of summaries of bug reports by using the sentence embedding. We used supervised learning technique to generate the summaries. In our proposed methodology the similarity between sentences is calculated by using sentence embedding. After preprocessing, the sentences are converted to vectors of real numbers by sentence embedding. K-mean cluster is used to cluster these sentences. Then we have to select one sentence per cluster. Sentence ranking is used to rank sentences per information they contain and select high rank sentences for summarization. We achieved improved rouge-1 and rouge-2 results than the previous state of the art systems for the bug report summary generation.