An IR-based Approach For Multilingual And Cross-lingual Ontology Matching Purposes

10^th November 2015, 13:00 Ashton Lecture Theater
Dr Mauro Dragoni
Fondazione Bruno Kessler (FBK)
Trento
Italy

Abstract

Ontology matching in a multilingual environment consists in finding alignments between ontologies modeled by using different languages. Such a research topic combines traditional ontology matching algorithms with the use of multilingual resources, services, and capabilities for easing either cross-lingual matching (where different ontologies are expressed by using different languages) and multilingual matching (where translations of all label are already available in all ontologies).
In this seminar, I will present our ontology matching approach able to work in both the cross-lingual (by exploiting an automatic machine translation service) and in the multilingual scenarios. The approach is based on the use of information retrieval (IR) techniques: ontologies are indexed through the inverted
index algorithm and candidate matches are found by querying such indexes. The approach has been evaluated on two different datasets: the OAEI MultiFarm benchmark and on set of domain-specific ontologies belonging to the agricultural and medical domains. Results demonstrating the viability of the approach from either the effective and efficient points of view.