Automated Recommendation of Related Model Elements for Domain Models

Domain modeling is an important activity in the early stages of software projects to achieve a common understanding of the problem area among project participants. Domain models describe concepts and relationships of respective application fields using a modeling language and domain-specific terms. Creating these models requires software engineers to have detailed domain knowledge and expertise in model-driven development. Collecting domain knowledge is a time-consuming manual process that is rarely supported in current modeling environments. In this paper, we describe an approach that supports domain modeling through formalized knowledge sources and information extraction from text. On the one hand, domain-specific terms and their relationships are automatically queried from existing knowledge bases. On the other hand, as these knowledge bases are not extensive enough, we have constructed a large network of semantically related terms from natural language data sets containing millions of one-word and multi-word terms and their quantified relationships. Both approaches are integrated into a domain model recommender system that provides context-aware suggestions of model elements for virtually every possible domain. We report on the experience of using the recommendations in various industrial and research environments.