Towards model reuse in low-code development platforms based on knowledge graphs

Abstract

Low-code Development Platforms (LCDP) are applications to provide fast full-stack application development. These platforms rely on models as their core artifacts in order to reduce the complexity of software development. Despite the growing need and importance of using models in low-code platforms, there is still limited support for model discovery and reuse, which leads to unnecessary repetitive work. Therefore, facilities for automated discovery and recommendation of relevant models and model fragments are desired. A prerequisite for producing relevant recommendations for adding new features to a model is a model repository equipped with an efficient query mechanism in combination with algorithms for processing the retrieved data. In this paper, we present an approach that enables model recommendations by initially converting and merging heterogeneous models into a homogeneous graph, which serves as the repository of our approach. Afterwards, the approach queries the repository for relevant matches and uses N-grams to produce recommendations for models under development. The current approach is demonstrated on the zAppDev LCDP, but conceptually it can be integrated on any LCDP and can be applied for the reuse of any graph-based model. We evaluated our approach by reconstructing four different models that do not exist in our repository with the support of our approach.