Non-Functional Requirement Detection Using Machine Learning and Natural Language Processing

Hazlina Shariff Et.Al

doi:10.17762/turcomat.v12i3.1171

Abstract

A key aspect of software quality is when the software has been operated functionally and meets user needs. A primary concern with non-functional requirements is that they always being neglected because their information is hidden in the documents. NFR is a tacit knowledge about the system and as a human, a user usually hardly know how to describe NFR. Hence, affect the NFR to be absent during the elicitation process. The software engineer has to act proactively to demand the software quality criteria from the user so the objective of requirements can be achieved. In order to overcome these problems, we use machine learning to detect the indicator term of NFR in textual requirements so we can remind the software engineer to elicit the missing NFR.We developed a prototype tool to support our approach to classify the textual requirements and using supervised machine learning algorithms. Survey wasdone toevaluate theeffectiveness of the prototype tool in detecting the NFR.

Highlights

Requirements elicitation is an important activity in the systems analysis and design process
They used a small database containing labelled examples to train the classifiers. They employed under- and oversampling strategies to handle the imbalanced classes in the dataset and cross-validated the classifiers based on the Support Vector Machine classifier algorithm
Non-Functional Requirement Detection Using Machine Learning And Natural Language Processing We evaluated the tool by doing closed-ended questionnaires with software engineers to validate the developed tool's ability to detect the non-functional requirements (NFR)

Summary

Introduction

Requirements elicitation is an important activity in the systems analysis and design process. To successfully address nonfunctional characteristics in these phases, it is essential to elicit and capture the NFR during the requirements engineering phase. It is normal sometimes when the user doesn't know to describe things they know. NFR is a tacit knowledge that user always faces the problem to express it especially during an elicitation process. Even when they know, user stories tend to be unclear, not precise and ambiguous and may lead the software developer to interpret in many ways because of the unfamiliarity of the NFR aspect.

Literature Review

Methodology

Results And discussion

Conclusion