NLP-based Auto Generation of Graph Database from Textual Requirements

NLP-based Auto Generation of Graph Database from Textual Requirements

연구 분야: Databases

학회: 2024 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)

Natural Language (NL) continues to be the preferred medium of expression for requirements, especially for multi-disciplinary systems such as the Flight Control System (FCS) of an Aircraft. The importance of a precise and unambiguous understanding of requirements for the project's success is established beyond question. This aspect is further amplified in the case of a safety-critical system such as FCS. UML diagrams from use case to activity to sequence diagrams are used to understand and analyze the requirements. While these diagrams have immense value, Graph Databases add new benefits for analyzing and managing requirements. Graph databases are gaining popularity due to their simplicity, scalability, and benefits of visualization. In this paper, we propose an NLP-based method for auto-generating a graph database for textual requirements. We explore the benefits of representing requirements as a graph database that complements the other diagrams and fills a few gaps, specifically in requirements allocation, identification of interdependencies, and change management. Though we use the Neo4j graph database management system and Cypher query language, the concepts are generic. The proposed approach for auto-generation works on requirements expressed in the EARS (Easy Approach to RequirementsSyntax) template. However, it can be extended for other templates or free-flowing text.