PhD Position F/M Computational approaches for knowledge graph mining and completion dealing wit[...]

INRIA; Valbonne; Sur place; EUR 40 000 - 60 000; Plein temps

Résumé du poste

Une institution de recherche en informatique recherche un candidat pour une thèse de doctorat axée sur le minage et la complétion de graphes de connaissances dans le cadre du projet MetaboLinkAI. Le candidat doit posséder un Master en Informatique et des compétences en normes du web sémantique, ainsi qu'une motivation pour la recherche. Ce poste inclut des avantages tels que des repas subventionnés, des congés annuels généreux, et des possibilités de télétravail après 6 mois.

Prestations

Repas subventionnés

Remboursement partiel des frais de transport

7 semaines de congés annuels + 10 jours de RTT

Possibilité de télétravail après 6 mois

Équipements professionnels disponibles

Événements sociaux, culturels et sportifs

Accès à la formation professionnelle

Couverture sociale

Qualifications

• Connaissances solides en normes et technologies du web sémantique.
• Compétences en interrogation et minage de graphes de connaissances.
• Motivation pour une recherche scientifique dans un contexte de science ouverte.

Responsabilités

• Développer des méthodes et outils pour le minage de graphes.
• Améliorer la complétude et la validité des graphes de connaissances.
• Intégrer un cadre de représentation et de raisonnement de l'incertitude.

Connaissances

Normes et technologies du web sémantique

Interrogation et minage de graphes de connaissances

Motivation pour la recherche scientifique

Compétences en développement

Compétences en rédaction

Anglais oral et écrit

Formation

Master en Informatique / Sciences informatiques

Description du poste

Contexte et atouts du poste

This PhD thesis takes place within the MetaboLinkAI ANR‑SNF project, which aspires to revolutionize the analysis and interpretation of metabolomics data through a multidisciplinary approach that combines a comprehensive knowledge graph hub (MetaKH) with cutting‑edge artificial intelligence (AI) and machine learning (ML) techniques. The project's main goals are to enhance the querying and ease of use of metabolomics data, improve research efficiency, and stimulate creativity in the field. These objectives are set to surpass current standards by creating an encyclopedic and expandable knowledge base, integrating advanced AI to handle the uncertainties of experimental data, and enabling a broader range of hypothesis testing and evaluation.

Within this project, we will focus on developing innovative methodologies and tools, such as graph mining methods, to enhance data interaction, analysis capabilities, and representation of uncertainty.

One distinctive peculiarity of metabolomics data (and thus MetaKH) is incompleteness, variable confidence and inherent uncertainty. Here, we adopt AI to enhance the completeness and reliability of the KG and to correctly account for uncertainty.

Mission confiée

Computational approaches for graph mining and completion

Because of the uncertain nature of metabolomics data and associated knowledge, MetaKH will be largely incomplete and partly incorrect. Therefore, it will be crucial to develop a comprehensive computational framework to enhance the quality, completeness and validity to eventually increase the quality of any processing using MetaKH. We propose to adapt heuristic methods and algorithms to discover/induce topological motifs, axioms (OWL), rules (SWRL or SPARQL) or shapes (SHACL) from knowledge graphs (TBox construction/refinement). These will account for the possible uncertainty of knowledge represented in the ABox ( WP3.2). Expert‑in‑the‑loop techniques will also be considered. We will design algorithms and data structures to allow KG queries at different levels of data granularity. The methods will exploit heuristics derived from expert knowledge in combination with semi‑succinct and, where needed, approximated data structures. In parallel, we will work on methods for knowledge graph completion, correction and enrichment, to enhance quality and content (ABox refinement). The developed methods will combine deductive reasoning (including analogic), SHACL validation, and link prediction and retraction based on KG embeddings. They will take into account the uncertainty of knowledge as defined in WP3.2. Evaluation will be done by measuring the improvement of KG completeness and validity, and the effectiveness of reasoning by corrupting the KG by adding/removing/perturbing some edges, applying completion/inference/querying, and assessing the impact in comparison with the original KG.

Dealing with (lack of) confidence in KGs

The objective is to develop and integrate a sophisticated framework into semantic web standards for formal representation and reasoning of uncertainty (both ontic and epistemic) in MetaKH, improving data confidence and decision‑making processes. Initially, we will review literature to identify adequate models to represent ontic uncertainty (certainly probability theory) and epistemic uncertainty (e.g. possibility theory, Dempster‑Shafer theory) adequate to represent mass spectrometry observations and metabolomic knowledge. Based on such models, we will propose extensions to Semantic Web standards to express uncertainty, provenance, and temporality metadata, facilitating richer data interpretation and trustworthiness. We will develop algorithms to integrate uncertainty in querying, deduction and embedding in KGs. We will establish criteria for using KGs based on uncertainty and provenance metadata, as well as other types of metadata, enabling users and agents to make informed decisions regarding trust and data application. Algorithms developed in WP3.1 will be extended to integrate uncertainty. Finally, we plan to implement mechanisms for evaluating KG completeness, validity, and reasoning under uncertainty, incorporating expert feedback and adapting methodologies based on provenance and other metaknowledge types.

Principales activités

This thesis will start with a state of the art of the different domains involved, in particular graph-based knowledge representation, KG mining, uncertainty representation and management in KG.

The PhD student is expected to first address computational approaches for MetaKH mining and completion, and then extend these approaches considering the inherent uncertainty of some knowledge in MetaKH, and of the mining approaches and their results.

Expected deliverables are:

• [D1] Heuristic methods, data structures and algorithms for KG querying and mining
• [D2] Methods and algorithms for KG completion
• [D3] Proposal of an extension of SW standards for uncertainty annotation
• [D4] Implementation of uncertainty annotation in MetaKH

References

• Ahmed El Amine Djebri. Uncertainty Management for Linked Data Reliability on the Semantic Web. PhD thesis, Université Côte d’Azur, 2022.
• Ahmed El Amine Djebri, Andrea G. B. Tettamanzi, and Fabien Gandon. Publishing uncertainty on the semantic web: Blurring the LOD bubbles. In Graph-Based Representation and Reasoning - 24th International Conference on Conceptual Structures, ICCS 2019, Marburg, Germany, July 1-4, 2019, Proceedings, volume 11530 of Lecture Notes in Computer Science, pages 42–56. Springer, 2019.
• Antonia Ettorre, Anna Bobasheva, Catherine Faron, and Franck Michel. A systematic approach to identify the information captured by knowledge graph embeddings. In WI-IAT ’21 : IEEE/WIC/ACM International Conference on Web Intelligence, Melbourne VIC Australia, December 14 - 17, 2021, pages 617–622. ACM, 2021.
• Rémi Felin. Evolutionary knowledge discovery from RDF data graphs. PhD thesis, Université Côte d’Azur, 2024.
• Rémi Felin, Catherine Faron, and Andrea G. B. Tettamanzi. A framework to include and exploit probabilistic information in SHACL validation reports. In The Semantic Web - 20th International Conference, ESWC 2023, Hersonissos, Crete, Greece, May 28 - June 1, 2023, Proceedings, volume 13870 of Lecture Notes in Computer Science, pages 91–104. Springer, 2023.
• Rémi Felin, Pierre Monnin, Catherine Faron, and Andrea G. B. Tettamanzi. An Algorithm Based on Grammatical Evolution for Discovering SHACL Constraints. In EuroGP 2024 - 27th European Conference on Genetic Programming, Genetic Programming – 27th European Conference, EuroGP 2024, Aberystwyth, United Kingdom, April 2024.
• Thu Huong Nguyen. Mining the semantic Web for OWL axioms. PhD thesis, University of Côte d’Azur, 2021.
• Andrea G. B. Tettamanzi, Catherine Faron-Zucker, and Fabien Gandon. Possibilistic testing of OWL axioms against RDF data. Int. J. Approx. Reason., 91 :114–130, 2017.

Compétences

The candidate must hold a Master degree in Informatics / Computer science and must demonstrate aptitudes or matches with most of the following aspects:

• Competencies and skills in Semantic Web standards and technologies
• Competencies and skills in querying and mining Knowledge Graphs
• High motivation for scientific research in an open science context
• Good development skills
• Writing skills and publication motivation
• Good English oral and writing skills

Soft skills

• Aptitude to work with others and engage in collaborations
• Autonomy and creativity
• Remote working capabilities (emails, collaborative tools, etc.)

Avantages

• Subsidized meals
• Partial reimbursement of public transport costs
• Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
• Possibility of teleworking (after 6 months of employment) and flexible organization of working hours
• Professional equipment available (videoconferencing, loan of computer equipment, etc.)
• Social, cultural and sports events and activities
• Access to vocational training
• Social security coverage

Rémunération

Gross Salary: 2300 €per month