Semi-automated development of conceptual models from natural language text

• Published in: Data & knowledge engineering 2020-05, Vol.127, p.101796, Article 101796
• Main Authors: Omar, Mussa, Baryannis, George
• Format: Article
• Language: English
• Subjects: Conceptual modelling; Information extraction; Natural language processing; Ontologies; Semi-structured data
• ISSN: 0169-023X; 1872-6933
• DOI: 10.1016/j.datak.2020.101796

The process of converting natural language specifications into conceptual models requires detailed analysis of natural language text, and designers frequently make mistakes when undertaking this transformation manually. Although many approaches have been used to partly automate this process, one of the main limitations is the lack of a domain-independent ontology that can be used as a repository for entities and relationships, thus guiding the transformation process. In this paper, a semi-automated system for mapping natural language text into conceptual models is proposed. The system, called SACMES, combines a linguistic approach with an ontological approach and human intervention to achieve the task. SACMES learns from the natural language specifications that it processes and stores the information that is learnt in a conceptual model ontology and a user history knowledge database. It then uses the stored information to improve performance and reduce the need for human intervention. The evaluation conducted on SACMES demonstrates that: (1) by using the system, precision and recall for users identifying entities of conceptual models is increased by 6% and 13%, respectively, while for relationships, increases are even higher, 14% for precision and 23% for recall; (2) the performance of the system is improved by processing more natural language requirements, and thus, the need for human intervention is decreased.