Offre en lien avec l’Action/le Réseau : DOING/– — –

Laboratoire/Entreprise : ISID and Vertigo teams at Centre d’études et de re
Durée : 6 mois
Contact : nada.mimouni@cnam.fr
Date limite de publication : 2024-03-01

Contexte :
In the field of cultural heritage, and painting in particular, the management of large collections has become increasingly complex over the years. Heritage data, including aspects such as names, creators, representations and images, have posed significant challenges for curators and researchers alike.
Semantic knowledge graphs have emerged as a promising approach to representing cultural heritage datasets. They provide a structured framework for integrating heterogeneous data sources, enabling comprehensive exploration and reasoning about cultural artworks and
their relationships. However, existing knowledge graphs are far from complete in this domain, and creating and populating semantic models for heritage data is a resourceintensive undertaking, requiring substantial human expertise. Knowledge graph completion (KGC) approaches have been proposed to enhance knowledge graphs by completing their missing connections. In this work, we aim to extend knowledge-graph completion techniques to this specific data domain, by leveraging both pre-trained language models and Graph Neural Networks (GNNs) to facilitate the efficient creation and extension of
knowledge graphs.

Sujet :
The work will be organized as follows:
– Related work: analysis of existing approaches on the use of pre-trained language models and GNNs to improve knowledge graph completion (KGC).
– Data: collection and creation of benchmarks to evaluate the models.
– Methodology: definition of a methodology for data preparation and knowledge graph enrichment.
– Interpretation and evaluation: carry out a quantitative assessment of the proposed methods for KGC, based on the created benchmarks, in order to establish their effectiveness in this context. An effort towards explaining these results should be made.

Profil du candidat :
A master degree in one or more of the following areas: machine learning, natural language processing, symbolic AI, semantic web.

Formation et compétences requises :
As a minimum requirement, the successful candidate should have:
• A master degree in one or more of the following areas: machine learning, natural language processing, symbolic AI, semantic web.
• Excellent programming skills (Java or Python)
• Excellent command of English
• Experience with machine learning and graphs

Adresse d’emploi :
Conservatoire National des Arts et Métiers Paris, 2 rue Conté, 75003

Document attaché : 202402231005_Internship_KGC_CNAM-list.pdf