Improving the clarity of requirements by generating RCI value using Machine Learning

Abstract

Requirement Engineering is a key phase in software development, which improves quality and maintainability of the software. In requirement engineering phase, there are several ways to measure the quality of the requirement, such as reliability, performance efficiency, security, maintainability, rate of delivery, testability and usability. Requirement Clarity Index (RCI) is a quality measure, that can be implemented in a system to reflect level of clarity each stakeholder has on the project requirements. In other words, RCI used to measure of having clear understanding on what stakeholder needs is essential for a successful software system delivery. Accurate identification of RCI will help reduce the ambiguity of requirements which reduce the rework, and improve maintainability. Measuring RCI value manually requires higher human involvement, which is expensive, time consuming and subjective. This research is intended to automate the requirement clarity index generation process using rule-based machine learning approach. Research has two main phases; (1) a text summarization phase and (2) requirement quality score analysis phase. Use of text summarization phase, natural language requirement details are summarized and key aspects of requirement details get extracted. In requirement quality score analysis phase, scoring is applied to summarized content, which is generated from text summarization phase, using identified quality factors from literature survey. Quality score for the factors returned from the quality score analysis phase. Using the generated quality score and manually computed RCI value, mapping table is created for rule based RCI generation approach. Mapping table contains range of metric scores and its related RCI value. Dataset of requirements undergoes into these phases and mapping table is generated. After generating the mapping table, requirements would be undergoes into the phases and quality score get computed. According to the quality score, particular range of quality score is mapped and respective RCI value is returned from the mapping table. With the help of large set of dataset, this research can produce more significant results for new requirements.