Personal page of Mariem Mahfoudh describing her research and publications.

Adaptation d’ontologies avec les grammaires de graphes typés : évolution et fusion

Advisors : Pr. Michel Hassenforder, Dr. Laurent Thiry, Dr. Germain Forestier

Funding : CCAlps European project

Manuscript : MahfoudhThesisMai2015.pdf

Abstract: Ontologies are a formal and explicit knowledge representation. They represent a given domain by their concepts and axioms while creating a consensus between a user community. To satisfy the new requirements of the represented domain, ontologies have to be regularly updated and adapted to maintain their consistency. The adaptation may take different forms (evolution, alignment, merging, etc.), and represents several scientific challenges. One of the most important is to preserve the consistency of the ontology during the changes. To address this issue, we are interested in this thesis to study the ontology changes and we propose a formal framework that can evolve and merge ontologies without affecting their consistency.

First we propose TGGOnto (Typed Graph Grammars for Ontologies), a new formalism for the representation of ontologies and their changes using typed graph grammars (TGG). A coupling between ontologies and TGG is defined in order to take advantage of the graph grammars concepts, such as the NAC (Negative Application Conditions), in preserving the adapted ontology consistency.
Second, we propose EvOGG (Evolving Ontologies with Graph Grammars), an ontology evolution approach that is based on the TGGOnto formalism that avoids inconsistencies using an a priori approach. We focus on OWL ontologies and we address both : (1) ontology enrichment by studying their structural level and (2) ontology population by studying the changes affecting individuals and their assertions.
EvOGG approach defines different types of ontology changes (elementary, composite and complex) and ensures their implementation by the algebraic approach of graph transformation, SPO (Single pushout).
Third, we propose GROM (Graph Rewriting for Ontology Merging), an ontologies merging approach that avoids data redundancy and reduces conflict in the merged result. The proposed approach consists of three steps: (1) the similarity search between concepts based on syntactic, structural and semantic techniques; (2) the ontologies merging by the algebraic approach SPO; (3) the global ontology adaptation with graph rewriting rules.

To validate our proposals, we have developed several open source tools based on AGG (Attributed Graph Grammar) tool. These tools were applied to a set of ontologies, mainly on those developed in the frame of the CCAlps (Creatives Companies in Alpine Space) European project, which funded this thesis work.

Keywords: Ontologies, Ontologies Evolution, Ontologies Merging, Typed Graph Grammars, Algebraic Graph Transformations, Simple Pushout (SPO)