Classification of Requirements, Bug Reports and Sample Clauses

Classification of Requirements, Bug Reports and. Feature Requests One topic in research is the classification of requirements into functional (FR) and non- functional (NFR) [42], [15], [35]. While there is some kind of an agreement on the definition of FRs, there is no clear definition of NFRs [16]. Both types of requirements are distinguished by whether they describe the functionality of the system (FR) or its properties and constraints (NFR) [45], [24]. This differentiation influences the practical use of requirements elicitation, documentation, and validation [12], [15]. In contrast to functional requirements, non- functional requirements are often specified later in the process and not explicitly managed [10], [11], [34], [36]. This led to a situation in which NFRs are often regarded as less important and delayed to a later point in the project, despite often influencing FRs [8], [11]. One approach to mitigate this issue is to automatically classify requirements into functional and nonfunctional. For example, Xxxxxxxxxx and Xxxxxx [42] employ a supervised machine learning algorithm that correctly classifies raw text requirements into FR and NFR 92% of the time. Similarly, distinguishing between bug reports and feature requests is also an important topic. Recently, research has focused on app stores—platforms to distribute mobile software where users can give feedback in terms of written text or through a rating system. With more than 5 million apps and hundreds of million users’ reviews, app stores are a rich source of information. By automatically mining information from app reviews, developers can leverage the users’ voice and improve their app [28]. For instance, developers could check the sentiment of reviews associated with certain features to understand how users perceive them [17]. On the other hand, due to their sheer number, many reviews are non-informative and need to be filtered out [9], [30]. In summary, the application of machine learning approaches to classification problems, interesting for requirements engineering, is showing good results. The OpenReq Requirements Intelligence will implement its classification functionalities based on the approaches reported in this section. Likewise, it will address the issues, here identified, of noisy data and bulkiness.