National and Kapodistrian University of Athens
Incomplete information in RDF databases with an emphasis on the geospatial domain
Abstract
dc:descriptionIncomplete information has been studied in-depth in relational databases and knowledge representation, but not much has been done in semantic web frameworks such as RDF. In this Thesis we propose an extension of RDF, called RDFi, for the representation and querying of incomplete information for property values. The RDFi framework is parameterized with a constraint language and serves for expressing partial knowledge about property values through the formulation of constraints. We give the syntax and formal semantics of RDFi and extend the compositional semantics of SPARQL to capture the new elements introduced in RDFi for the capturing of incomplete information. We show two fragments of SPARQL that can be used to evaluate queries over RDFi in a semantically correct way and only with respect to a subset of their answers corresponding to the so-called certain answers. Our work around RDFi completes by giving an algorithm for the computation of certain answers and studying the data complexity of the associated decision problem named certainty. The rest of our work is more practical. Motivated by the rapid increase in the size of linked geospatial data published on the web, we give a strong emphasis on the geospatial domain. We point out that the recent OGC standard GeoSPARQL for the querying of geospatial information provides a rich language, but at the same time it covers a number of weaknesses pertaining to the handling of incomplete information that leading to incomplete answers. In this context, we demonstrate the expressivity of the RDFi framework showing that it can be employed to give semantics to GeoSPARQL. We continue by discussing possible ways through which these problems can be amended using constraint propagation techniques developed in the area of qualitative spatial reasoning and then presenting an implementation as an extension to the system Strabon as well as a corresponding evaluation using linked geospatial datasets. Prompted by the negative results of this evaluation we then study how consistency checking of very large real-world networks, such as the linked geospatial datasets, can be done efficiently.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2015
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Nikolaou, Charalampos
- Νικολάου, Χαράλαμπος
Subjects
dc:subject × 15Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/38588