A Zero-Shot Learning Approach to Classifying Requirements: A Preliminary Study

Context and motivation: Advances in Machine Learning (ML) and Deep Learning (DL) technologies have transformed the field of Natural Language Processing (NLP), making NLP more practical and accessible. Motivated by these exciting developments, Requirements Engineering (RE) researchers have been experimenting ML/DL based approaches for a range of RE tasks, such as requirements classification, requirements tracing, ambiguity detection, and modelling. Question/problem: Most of today's ML/DL approaches are based on supervised learning techniques, meaning that they need to be trained using annotated datasets to learn how to assign a class label to examples from an application domain. This requirement poses an enormous challenge to RE researchers, as the lack of requirements datasets in general and annotated datasets in particular, makes it difficult for them to fully exploit the benefit of the advanced ML/DL technologies. Principal ideas/results: To address this challenge, this paper proposes a novel approach that employs the Zero-Shot Learning (ZSL) technique to perform requirements classification. We build several classification models using ZSL. We focus on the classification task because many RE tasks can be solved as classification problems by a large number of available ML/DL methods. In this preliminary study, we demonstrate our approach by classifying non-functional requirements (NFRs) into two categories: Usability and Security. ZSL supports learning without domain-specific training data, thus solving the lack of annotated datasets typical of RE. The study shows that our approach achieves an average of 82% recall and F-score. Contribution: This study demonstrates the potential of ZSL for requirements classification. The promising results of this study pave the way for further investigations and large-scale studies. An important implication is that it is possible to have very little or no training data to perform requirements classification. The proposed approach thus contributes to the solution of the long-standing problem of data shortage in RE.