Towards Inductive Query-Relaxation for RDF

Presented at: 9th Extended Semantic Web Conference (ESWC2012)

by Aidan Hogan, Marc Mellotte, Gavin Powell, Dafni Stampouli

In this paper, we argue that query relaxation over RDF data is an important but largely overlooked research topic: the Semantic Web standards allow for answering crisp queries over crisp data, but what of use-cases that require approximate answers for structured queries over RDF data? We introduce one such use-case extracted from an EADS project that aims to fuse together intelligence information for police post-incident analysis. The concrete application for query relaxation involves matching (possibly vague) descriptions of entities involved in crimes to structured descriptions thereof in the database. Here, the core research questions are: (i) how can we formalise potentially vague structured queries in a generic manner; (ii) how can we support approximate, structured query-answering over RDF? We first discuss the use-case, formalise the problem, and survey current literature for possible approaches. Next, we present a proof-of-concept framework for enabling relaxation of structured entity-lookup queries, evaluating different distance measures for performing relaxation. We argue that, beyond our specific use-case, query relaxation is important to many potential use-cases for Semantic Web techniques, and worthy of further attention.

Keywords: cooperative systems, query answering, query relaxation

Resource URI on the dog food server:

Explore this resource elsewhere: