Abstract
With the growing awareness of the effects of requirements on software development processes, requirements engineering is increasingly becoming an important field of focus in software engineering research. Many studies show that failures in understanding and classifying requirements are the main reasons for exceeding project costs and allocated time, which in turn may cause the failure of a project. Successful software systems development requires consistent and classified requirements. The classification of requirements represents an early but critical phase in the requirements analysis stage. While the literature sheds light on distinctions between different types of requirements, the detection of such differences in practice is not always an easy task. This paper provides an overview of requirements classification, presents some of the existing research studies on requirements classification, and discusses their limitations in terms of yielding suggestions for improvement. Additionally, this work takes a different approach to address requirements classification. It proposes a semantic model to classify requirements automatically, using a hybrid artificial intelligence approach. In addition, the paper discusses evaluation methods for each part of the proposed model.