Constructing Formal Specification Models from Domain Specific Natural Language Requirements
Jun Hu, Jiancheng Hu, Wenxuan Wang, Jiexiang Kang, Hui Wang, Zhongjie Gao · 2020
One important way to improve the quality of safety-critical software is to produce a good software requirement satisfying several key properties, such as: integrity, consistency, and well organized, etc. Our work is based on airborne software domain, and propose a framework to translate the software requirements, which are itemized with domain natural language in avionics, effectively into a formal specification model VRM (Variable Relation Model), which has table-style structures with formal semantics. Firstly, considering avionics domain characteristics, a domain concept library is established including different types of variables and concepts. Then, a set of domain-oriented requirements templates are defined, such as: general event/condition, display event/condition, etc. According to VRM model element semantics, three types model construction algorithms are designed to complete the translation automatically. And in the case study, the Engine Indication and Crew Warning System (EICAS) was selected to show how to construct formal models from natural language requirements.