Identification of non-functional requirements in textual specifications: A semi-supervised learning approach

Early detection of non-functional requirements (NFRs) is crucial in the evalua-tion of architectural alternatives starting from initial design decisions. The ap-plication of supervised text categorization strategies for requirements expressed in natural-language has been proposed in several works as a method to help analysts in the detection and classification of NFRs concerning different aspects of software. However, a significant number of pre-categorized requirements is needed to train supervised text classifiers, which implies that analysts have to manually assign categories to numerous requirements before being able of accurately classifying the remaining ones. In this paper, we propose a semi-supervised approach which bases classification on a reduced number of catego-rized requirements by taking advantage of the knowledge provided by uncat-egorized ones, as well as certain properties of text. Thus, a small number of requirements, possibly identified by the requirement team during the elicitation process, enables learning an initial classifier for NFRs, which could successivel