FACT-from actual to conceptual tie-ins: A multi-level knowledge graph structured on context and semantics of software artefacts

Abstract

Knowledge Graphs have increasingly become the preferred approach to complex problems that involve low-level business data. It is well known that Google, LinkedIn, Facebook, Twitter - all have a knowledge graph at their core. In this paper, we propose: i) A knowledge graph, called FACT, which is designed for applications in software engineering problems. ii) A novel system which can automatically populate FACT. Example applications of FACT include verifying design decisions, recommending software elements to reify design decisions, and so on. Vertices of FACT represent items such as a software design concept, a concrete software element which reifies a concept. An edge represents the relationship that may exist between the vertices. The design of FACT as well as the system used to populate it has been validated a) at micro-level by verifying or proving the correctness of the individual components of FACT, and b) at macro-level by experimentally ascertaining the correctness of scenario-based inferences derived from the knowledge contained in FACT.