A recommender system to assist conceptual modeling with UML

Abstract

This paper explores the understudied field of conceptual modeling assistance. More specifically, we focused on the design and application of recommender systems as software assistants for conceptual modeling. Prior work on such systems has shown that trust plays a key role in the acceptance and exploitation of such systems. Consequently, as a starting point of our research, we applied established methods for constructing multi-criteria recommender systems (MCRS) to conceptual modeling in a way which could foster the emergence of trust. Finally, we chose supervised-learning techniques to refine and customize the recommendations generated by these systems. To help us determine the feasibility and practicality of our approach, we designed and implemented a prototype system that assists conceptual modeling with UML. Our system currently recommends class attributes when constructing UML class diagrams. A preliminary evaluation of this tool indicated a strong match between the recommendations provided by our system and personal choices made by the participants.