Mining Software Requirements From Turkish Texts: Techniques and Challenges

Abstract

Extraction of software requirements from natural language texts becomes more important in requirements engi-neering for the identification o f stakeholder needs hidden in big documentation. However, it ensures that the system goals will be completely understood. This step is principal for improving communication, refining documentation, and providing a clear set of requirements that guide software development and quality as-surance. The complexity and ambiguity of natural languages are significant, and mostly result in misinterpretations or incomplete requirements. The relevant information is usually scattered over several documents and communication channels, thus accurate capture remains still difficult. Extracting requirements from Turkish texts adds further challenges due to its unique features, such as high level of agglutination. This linguistic complexity makes natural language processing (NLP) tasks like tokenization, morphological and syntactic analysis more challenging. The great problem is even further amplified b y t he fact that t here are very few comprehensive NLP tools and resources available for the Turkish language. This paper reviews NLP methods used for software requirement extraction, emphasizing challenges and techniques that are applicable to Turkish texts in general and, specifically, the richness of morphology and word order flexibility and the tight-associated inherent ambiguities in this language. Moreover, we cover current studies in this area and describe the available libraries, tools, and resources for NLP in Turkish, pointing out the limitations and possible lines of future research. © 2024 IEEE.