Detecting Bad Smells in Industrial Requirements Written in Natural Languages

Sammanfattning

A key factor in creating software of good quality is that the requirements for the project being developed are as unambiguous and clear as possible, so the developers will be able to develop the product quickly and effectively. So, there is a need for tools that help requirements engineers create quality requirements. The attributes that define a poorly written requirement are called bad smells. In this thesis we investigate the NALABS tools bad smell detecting capabilities when analyzing industrial requirements. First, we performed a literature study to investigate what types of bad smells exist for requirements and how they were specified. After that we used a case study to examine how many smells and of what categories the NALABS tool detects, when it analyzes industrial requirements. Lastly, we used a small experiment to examine how accurately NALABS detects smells, by designing a simple console application that counted instances of bad smell words in a set of keywords that were from the NALABS tool. The results we gathered gave us an indication that NALABS detects bad smells in all the categories of bad smells that are implemented in it, to a varying degree. Through this thesis we hope to extend the knowledge about bad requirements smells, clarify what attributes of a requirement might be a bad smell, and investigate to what degree the NALABS tool can detect bad smells in industrial requirements.