A content-based approach for recommending UML sequence diagrams

Abstract

Software engineers usually have to face a large space of choices during the development process, including libraries/ APIs, frameworks and UML models, which undermines their ability in finding the ones that best fit their needs. Recommender Systems appear as a solution to this problem since they have been applied successfully in other domains that suffer from similar issues. In this paper we propose to recommend UML Sequence Diagrams, a popular software artifact in many development processes, as an attempt to mitigate this problem. Our approach consists of: (i) a suitable representation of the users' information needs and sequence diagrams' content; and (ii) two content-based recommendation algorithms to recommend sequence diagrams that match the users' preferences. We performed a study with computer science subjects, where we generated recommendations with (ii) and measured the users' satisfaction upon these recommendations. Our preliminary results show that both algorithms are able to provide accurate recommendations.