Study of Various Classifiers for Identification and Classification of Non-functional Requirements

• Published in: Computational Science and Its Applications - ICCSA 2018 Vol. 10964; pp. 492 - 503
• Main Authors: Tóth, László, Vidács, László
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Classification; Natural language processing; Non-functional requirements; Requirements elicitation
• ISBN: 9783319951737; 3319951734
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-95174-4_39

Identification of non-functional requirements in an early phase of software development process is crucial for creating a proper software design. These requirements are often neglected or given in too general forms. However, interviews and other sources of requirements often include important references also to non-functional requirements which are embedded in a bigger textual context. The non-functional requirements have to be extracted from these contexts and should be presented in a formulated and standardized way to support software design. The set of requirements extracted from their textual context have to be classified to formalize them. This task is to be accomplished manually but it can be very demanding and error-prone. Several attempts have been made to support identification and classification tasks using supervised and semi-supervised learning processes. These efforts have achieved remarkable results. Researchers were mainly focused on the performance of classification measured by precision and recall. However, creating a tool which can support business analysts with their requirements elicitation tasks, execution time is also an important factor which has to be taken into account. Knowing the performance and the results of benchmarks can help business analysts to choose a proper method for their classification tasks. Our study presented in this article focuses on both the comparison of performances of the classification processes and their execution time to support the choice among the methods.