Evaluating and Improving Software Quality Using Text Analysis Techniques - A Mapping Study

Improvement and evaluation of software quality is a recurring and crucial activity in the software development life-cycle. During software development, software artifacts such as requirement documents, comments in source code, design documents, and change requests are created containing natural language text. For analyzing natural text, specialized text analysis techniques are available. However, a consolidated body of knowledge about research using text analysis techniques to improve and evaluate software quality still needs to be established. To contribute to the establishment of such a body of knowledge, we aimed at extracting relevant information from the scientific literature about data sources, research contributions, and the usage of text analysis techniques for the improvement and evaluation of software quality. We conducted a mapping study by performing the following steps: define research questions, prepare search string and find relevant literature, apply screening process, classify, and extract data. Bug classification and bug severity assignment activities within defect management are frequently improved using the text analysis techniques of classification and concept extraction. Non-functional requirements elicitation activity of requirements engineering is mostly improved using the technique of concept extraction. The quality characteristic which is frequently evaluated for the product quality model is operability. The most commonly used data sources are: bug report, requirement documents, and software reviews. The dominant type of research contributions are solution proposals and validation research. In our mapping study we identified 81 relevant primary studies. We pointed out research directions to improve and evaluate software quality and future research directions for using natural language text analysis techniques in the context of software quality improvement.