Logic-Based Integration of Query Answering and Knowledge Discovery

Abstract

There is currently great interest in integrating knowledge discovery research into mainstream database systems. Such an enterprise is nontrivial because knowledge discovery and database systems are rooted in different paradigms, therefore foundational work needs to be carried out and a candidate unified syntax and semantics needs to be proposed. Elsewhere we have indeed carried out such foundational work and used it to propose a unified syntax and semantics for integrating query processing and knowledge discovery. We refer to the resulting class of database systems as combined inference database systems (CIDS), since they are a class of logic-based databases and the integration is anchored by a view of query answering as deductive inference and of knowledge discovery as inductive inference. The most important novel capability of CIDS is that of evaluating expressions which seamlessly compose query answering and knowledge discovery steps. This gives rise to increased flexibility, usability and expressiveness in user interactions with database systems, insofar as many relevant and challenging kinds of information needs can be catered for by CIDS that would be cumbersome to cater for by gluing together existing, state-of-the-art (but, syntactically and semantically, heterogeneous) components. In this paper, we provide an overview of CIDS, then we introduce two motivating applications, we show how CIDS elegantly support such challenging application needs, and we contrast our work with other attempts at integrating knowledge discovery and databases technology.