• PhD Student:
  • Truong Thanh Ma
  • Funding : Artois
  • PhD defended on :
  • Dec 6, 2022 • Salle des thèses

The subject of the thesis is ontology merging, an approach for integrating various ontology sources into a unique one that handles emerging conflicts. This dissertation takes inspiration from the belief merging theory for merging ontologies to produce a consistent and unified knowledge base. We propose the three following contributions. ​ ​The first one is a semantic-based merging approach. In particular, a model-based merging strategy mainly focuses on handling semantic conflicts. A semantic conflict, which is not necessarily logical, is knowledge represented in many different or opposite ways. We also propose a formal model characterization and show the method’s effectiveness with an experimental evaluation.

​The second approach proposes a new framework to merge open-domain terminological knowledge. It leverages RCC5, a formalism for representing regions in a topological space and reasoning about their set-theoretic relationships. We propose a faithful translation of terminological knowledge from conflicting sources into region spaces. Here, we merge knowledge bases in this space and translate the outcome into the input sources’ language. Our technique uses RCC5’s expressivity and flexibility to deal with contradictory knowledge.

​The last contribution is a framework for evaluating ontology merging operators. The primary strategy starts with an original ontology to create noisy ontologies as datasets and use them to assess the merging operators. Then, we analyze merging operators’ computation time effectiveness and ability to cover the original ontology. Finally, we experimented with practical ontologies to evaluate the merging operators.

Thesis committee

Reviewers:

​- Eduardo FERMÉ, University of Madeira ​- Odile PAPINI, IS-CNRS, Aix Marseille Université, Marseille

Examiners:

  • Salem BENFERHAT, CRIL, CNRS & Artois University
  • Fatiha ​ SAÏS, LISN, Université Paris Saclay
  • Nicolas SCHWIND, AIST - Artifical Intelligence Research Center

Supervision:

  • Sébastien KONIECZNY, CNRS
  • Ivan VARZINCZAK, LIASD, Université Paris 8
  • Zied BOURAOUI, CNRS & Artois University