Context-based ontology matching: foundations and experiments

The semantic web relies on the expression of formalized knowledge on the web (in languages like RDF). Like the web, the semantic web will have to be distributed and heterogeneous. As such, the integration of resources found on the semantic web is one of its main problems. For contributing solving this problem, data is expressed in the framework of ontologies (theories describing the vocabulary used for expressing data). However, ontologies themselves can be heterogeneous and have to be reconciled.

One way to reconcile ontologies is to find the correspondences between their entities. This is called ontology matching [1] and the resulting set of correspondences is called an alignment. Ontology matching provides a set of correspondences supposed to hold between entities of two ontologies provided as input. A correspondence is defined by a relation between two entities (which can be classes, individuals, properties, termes or formulas involving such terms).

Context-based ontology matching is a relatively new way of finding relationships between two ontologies. It works by taking advantage of intermediate ontologies to which the two ontologies to be matched are connected. Then, the relationships between connected entities are composed to provide a correspondences between these entities. There have been several systems implementing context-based matching in specific ways. In particular, some work have considered using one intermediate domain-specific ontology in medicine [2], some systems use upper-level ontologies, while Scarlet [3] was designed to use all the ontologies on the web.

We have started to design a framework which offers much flexibility in the way context-based matching can be achieved. This framework reveals that there are many options in designing context-based matching algorithms:

• What kind of intermediate ontologies are used: all ontologies, domain specific ontologies, upper-level ontologies?
• How to choose intermediate ontologies: based on available alignents, based on pair of entities, maximum common entities in the two ontologies?
• How to match entities: label equality? label synonyms? elaborate matcher?
• How to aggregate different results: confidence, disjunctive relation intersection, result majority vote?
• How to traverse composition graphs: always use them, use the first that works, use them if everything else fails?
• What kind of composition is used: functional, subsumption, relational, ontology-based?

The goal of this post-doctoral position is to investigate this framework on the theoretical and/or experimental level. This can be by implementing and experimenting this framework and to experimentally determine what are the most efficient combinations of parameters and why. This can also be in precisely defining the algorithms and showing how they approach an ideally complete algorithm.

References:

• [1] Jérôme Euzenat, Pavel Shvaiko, Ontology matching, Springer, Heildelberg (DE), 2007
• [2] Zharko Aleksovski, Using background knowledge in ontology matching, PhD thesis, Vrije Universiteit Amsterdam, 2008
• [3] Marta Sabou, Mathieu d'Aquin, Enrico Motta, Using the semantic web as background knowledge for ontology mapping, Proc. 1st ISWC Ontology Matching workshop, Athens (GA US), 2006

Qualification: PhD or equivalent in computer science. Knowledge in knowledge representation, ontologies or ontology matching is welcome.

Researched skills:

• Interaction with other researchers.
• Taste for experiments and/or theory.
• Autonomous researcher.
• Writing capabilities.
• Innovative.

Hiring date: as soon as possible.

Place of work: The position is located at INRIA Grenoble Rhône-Alpes, Montbonnot (near Grenoble, France) a main computer science research lab, in a stimulating research environment. Research will be carried out in the Exmo team under the supervision of Jérôme Euzenat. It will require the involvement of the candidate in related projects.

Duration: 12 months (extendible to longer period)

Salary: 2357 EUR/month (before charges)

Contact: For further information, contact Jerome:Euzenat#inrialpes:fr.
Some more administrative information is available (but do not hesitate to contact us any way since we are more flexible).

Procedure: Visit INRIA's presentation (including FAQ and forms).
Note that, however, the procedure can be faster if judged necessary. In addition, send the same files to the contact above.

INRIA Priority research themes:

• Designing and mastering the future network infrastructures and communication services platforms/Concevoir et maîtriser les futures infrastructures des réseaux et des services de communication;
• Developing multimedia data and multimedia information processing/Développer le traitement des informations et données multimédia.