Section: New Results

Ontology matching and alignments

We pursue our work on ontology matching and alignment support [5] , [12] with contributions to evaluation and alignment semantics.

Evaluation

Participant : Jérôme Euzenat.

Since 2004, we run the Ontology Alignment Evaluation Initiative (oaei ) which organises evaluation campaigns for assessing the degree of achievement of actual ontology matching algorithms [2] .

This year, we ran the oaei 2013 evaluation campaign [7] . It offered 8 different test sets (7 of which under the seals platform). This issue brought the following results:

• Once again, more participants than ever (23);

• Most ontology matchers running on the SEALS platform (20);

• Increased performances in terms of precision and recall;

• Matchers are now faster and more scalable. There are also more matchers using networked resources.

We used again the our generator for generating new version of benchmarks [4] . The Alignment api was used for manipulating alignments and evaluating results.

A novelty of this year was the evaluation of interactive systems, included in the seals client. It brings interesting insight on the performances of such systems and should certainly be continued.

The participating systems and evaluation results were presented in the 8th Ontology Matching workshop, that was held in Sydney, Australia [13] . More information on oaei can be found at http://oaei.ontologymatching.org/ .

Algebras of relations in alignments

Participants : Armen Inants [Correspondent] , Jérôme Euzenat.

We had previously shown that algebras of relations between concepts can be used for expressing relations in alignments. We have worked this year as extending them in two ways.

We increased the expressiveness of relations between concepts, not restricting the algebra to necessarily non empty concepts. This describes all taxonomical (as opposed to mereological) relation algebras, i.e., all those relations that have been used by matchers so far.

We also dealt with relations among different kinds of entities – individuals or concepts. For this, relation algebra structures are considered in an arbitrary one- or many-sorted logical theory. We established a sufficient condition for a set of dyadic formulas in a first-order theory to generate a relation algebra. This result is extended to many-sorted theories by means of Schröder categories.

This work is part of the PhD of Armen Inants.