Improving the quality of software issue report descriptions in Turkish: an industrial case study at Softtech

Official URL: https://dx.doi.org/10.1007/s10664-023-10434-4

Issue reports are an important part of the software development process. They help developers identify and fix problems in their code. However, problems described in these reports often lack important information, such as the Observed Behavior (OB), Expected Behavior (EB), and Steps to Reproduce (S2R). This can lead to valuable developer time being wasted on gathering the relevant information. This study aims to address this issue by developing a tool that guides reporters in providing the necessary information in an industrial setting. The study is conducted at Softtech, a software subsidiary of the largest private bank in Turkey. The proposed approach is developed for issue reports written specifically in Turkish language. It is motivated by the need for issue report classification tools that can handle the unique characteristics of the Turkish language, such as the presence of many compound words. We first manually analyze and label 1, 041 issue reports for the existence of OB, S2R, and EB, and then present the specific patterns we found describing the related information. Next, we use morphological analysis to extract keywords and suffixes, and then use them for classification with a machine learning based approach. In addition, we conduct a feasibility study to assess the potential of using large language models for issue report classification tasks as a direction for future research. The results indicate that the tool using the machine learning-based approach can be used to guide in improving the quality of issue reports at Softtech, thereby saving valuable developer time.