Towards Efficient Use Case Modeling with Automated Domain Classification and Term Recommendation

Abstract

In requirements engineering, it takes significant time to specify requirements of various formats. Quality of specified requirements has direct impact on subsequent activities of software development, such as analysis and design. Motivated by this, in the paper, we aim to reduce effort required for specifying use case models and meanwhile improve their quality (in terms of consistency and correctness, for instance). Specifically, we investigate how to automatically classify a domain and recommend domain terminologies with natural language processing and information retrieval techniques, in the context of applying Restricted Use Case Modeling (RUCM) for developing use case models in natural language. To evaluate our approach (named RUCMBot), we evaluate it with seven subject systems. Results indicate that RUCMBot can help RUCM users by recommending domain terms with the accuracy being 0.6 in terms of F-score, on average. Moreover, RUCMBot is able to 100% correctly classify domains. RUCMBot also demonstrates its capability of constructing the domain terminology dictionary, and subsequently enhancing its recommendation accuracy along with the continuous use of RUCM for use case modeling.