Querying Incomplete Geospatial Information in RDF

Incomplete information has been studied in-depth in relational databases and knowledge representation. It is also an important issue in Semantic Web frameworks such as RDF, description logics, and OWL 2. In [6], we introduced RDF, an extension of RDF for representing incomplete information using constraints. We defined a semantics for RDF and studied SPARQL query evaluation in this framework. Given the current interest in publishing geospatial datasets as linked data (e.g., by Ordnance Survey in the UK), RDF is an excellent framework for encoding, possibly incomplete, qualitative and quantitative geospatial information which is found in these published datasets. RDF is also interesting because when the constraint language used can express the topological relations of RCC-8 [8], the recent OGC standard GeoSPARQL [7] for querying geospatial information expressed in RDF, becomes a special case of RDF. In this paper, we propose the problem of implementing an efficient query processing system for incomplete temporal and geospatial information in RDF as a challenge to the SSTD community. For the case of incomplete temporal information in relational databases, two such systems have been implemented in the past [1,3], but their query languages are rather limited. There is also the paper [2] which studies a closely related problem, temporal relationships in databases, but which has no related implementations. Finally, the knowledge representation language Telos [5] allows the modelling of incomplete temporal knowledge but well-known implementations such as ConceptBase have not implemented this functionality. To the best of our knowledge, no such relational database system or RDF store exists for the geospatial case, although there are some description logic reasoners that come close in terms of functionality [11]. In the rest of the paper, we present the RDF framework and outline the hard problems that have to be solved if such a query processing system is to become a reality. As in the theoretical foundation of [6], we concentrate on geospatial information only.