Analyzing Quality of Software Requirements; A Comparison Study on NLP Tools
Afrah Naeem, Zeeshan Aslam, Munam Ali Shah · 2019
For analyzing the quality of software requirements and performing natural language (NL) processing, only few of many tools are currently adopted by the industry. Most of these tools are no longer available and are obsoleted due to their complexity, less accuracy and high cost. The only few remaining tools exhibit either optimum functionality or at sub-optimal level. Some tools are limited to processing NL whereas the rest check for grammatical mistakes. First, this paper performs comparison on five existing tools named as QVscribe, QuOD, Innoslate, RAT and RQA in which we compare features, functionality and quality indicators addressed by these tools. Second, we propose a new tool named as Requirement Assessment Tool (RAT) for analyzing quality of individual requirements and specification document. Finally, check the performance of these tools for automatically detecting ambiguity, deviations and quality affecting defects on the basis of real-world natural language requirements set i.e. golden standard data. The results of these tools concluded that RAT gives more accurate results for detecting defects in all categories except looking for negative sentences.