Test case information extraction from requirements specifications using NLP-based unified boilerplate approach

• Main Authors: Lim, Jin Wei, Chiew, Thiam Kian, Su, Moon Ting, Ong, SimYing, Subramaniam, Hema, Mustafa, Mumtaz Begum, Chiam, Yin Kia
• Format: Article
• Published: Elsevier 2024
• Subjects: QA75 Electronic computers. Computer science; QA76 Computer software
• Online Access: http://eprints.um.edu.my/45561/; https://doi.org/10.1016/j.jss.2024.112005

Automated testing which extracts essential information from software requirements written in natural language offers a cost-effective and efficient solution to error-free software that meets stakeholders' requirements in the software industry. However, natural language can cause ambiguity in requirements and increase the challenges of automated testing such as test case generation. Negative requirements also cause inconsistency and are often neglected. This research aims to extract test case information (actors, conditions, steps, system response) from positive and negative requirements written in natural language (i.e. English) using natural language processing (NLP). We present a unified boilerplate that combines Rupp's and EARS boilerplates, and serves as the grammar guideline for requirements analysis. Extracted information is populated in a test case template, becoming the building blocks for automated test case generation. An experiment was conducted with three public requirements specifications from PURE datasets to investigate the correctness of information extracted using this proposed approach. The results presented correctness of 50 % (Mdot), 61.7 % (Pointis) and 10 % (Npac) on information extracted. The lower correctness on negative over positive requirements was observed. The correctness by specific categories is also analysed, revealing insights into actors, steps, conditions, and system response extracted from positive and negative requirements.