research

My research lies at the crossroads of symbolic and neural artificial intelligence, with a focus on extracting, representing and exploiting knowledge through knowledge graphs, natural language processing and machine learning.

See also: publications, projects, events, talks.

Research axes

Knowledge Graphs and GNNs

From the characterisation of semantic relations to graph neural networks.

  • Extraction and classification of semantic relations (SemEval-2018 Task 7, EMNLP 2017, IDA 2016)
  • Construction of large-scale knowledge graphs (AI-KG, ISWC 2020; CS-KG, ISWC 2022)
  • Geometric constraints in GNNs, such as edge curvature and Delaunay triangulation, against over-smoothing and over-squashing (ICML 2024; Data & Knowledge Engineering 2025)

Open science and automatic access to scientific literature

  • Construction and completion of scientific knowledge graphs (AI-KG, CS-KG; SCICERO, Knowledge-Based Systems 2022; research hypothesis generation, Knowledge-Based Systems 2025)
  • Automatic indexing of scientific documents (NLDB 2022)
  • Citation prediction with transformers (Information Processing & Management 2024)

Reliability and efficiency of large language models

  • Detection of AI-generated text in scientific publications (LREC 2022; NLDB 2023)
  • Predicting and localising memorization in fine-tuned LLMs (ECAI 2025; AAAI 2026)
  • Multilingual factual hallucination detection (MUCH benchmark, LREC 2026)
  • Interpretability of LLM classifiers with sparse autoencoders (Findings of EACL 2026)
  • Computational efficiency of vision-language models (PACT, CVPR 2025)

Analysis of online content: figurative language, opinions and disinformation

  • Detection of figurative language, irony and humour (Data & Knowledge Engineering 2012; IEEE Computational Intelligence Magazine 2019)
  • Identification of malicious online behaviour: bots, content spammers, trolls (EJASA 2020; TexTrolls; ECNLP 2024)
  • Classification of multimodal hateful content (CLEF 2024)

PhD students

Université Sorbonne Paris Nord (LIPN)

  • Tian Fang (October 2025 - ): emotion recommendation for personalised emotional feedback. Director, co-supervised with Gaël Guibon
  • Hugo Attali (September 2023 - July 2026): Graph Neural Networks for the automatic correction of Knowledge Graphs. Co-supervised with Nathalie Pernelle
  • Vijini Liyanage (March 2021 - May 2024): Knowledge-based detection of automatically generated text in scientific publications. Co-supervised with Adeline Nazarenko
  • Simon David Hernández Pérez (February 2016 - December 2019): Topic detection and similarity in scientific literature. Co-supervised with Thierry Charnois
  • Ehab Hassan (January 2013 - May 2017): Event-based recognition of lived experiences in user reviews. Co-supervised with Aldo Gangemi

École Polytechnique (LIX)

  • Jérémie Dentan (September 2023 - ): memorization and security of language models. Co-supervised with Sonia Vanier
  • Mohamed Dhouib (September 2023 - ): vision-language models. Co-supervised with Sonia Vanier
  • Hadi Abdine (December 2020 - February 2024): transformer-based language models for specialised domains. Co-supervised with Michalis Vazirgiannis
  • Nabil Moncef Boukhatem (September 2020 - December 2024), CIFRE thesis: semantic interpretation of spoken French and English technical dialogues. Co-supervised with Leo Liberti

Other institutions

  • Maxime Pegane (2024 - ), CY Tech CIFRE thesis, co-supervised with Besma Zeddini (SATIE, CY Cergy Paris Université)

Visiting PhD students

  • Vicent Ahuir Esteve (Universidad Politécnica de Valencia), 2024: relation extraction with seq2seq models
  • Farid Bagheri (Università di Cagliari), 2023-2024: reviewer recommendation for journals and conferences
  • Juan Luis García-Mendoza (INAOE Puebla), 2022: noise reduction in distant supervision
  • Bilal Ghanem (Universidad Politécnica de Valencia), 2019: detection of malicious bots on social media
  • Danilo Dessì (Università di Cagliari), 2019: graph-based analysis of scientific literature
  • Laura Di Rocco (Università di Genova), 2018: knowledge-based geotagging of tweets
  • Farideh Tavazoee (Università di Cagliari), 2017-2018: statistical analysis of social media

Master and bachelor students

Bachelor theses, École Polytechnique

  • Kingshuk Gupta (2026): adapting the Ghost-in-transformer methodology to detect hallucination spans best X BSc Thesis prize
  • Marc Saadé (2026): GNNs for low-CO2 route prediction in maritime shipping
  • Jay Ansh Bindra (2026): deep learning for the prediction of collisions in dense marine traffic
  • Andrea Foffani Pifarré (2024): LLM-based extraction of semantic triples from text
  • Nguyen Ngoc Tu (2024): detecting LLM-generated content in text
  • Mija Pilkaite (2022): classification of weather events with the ESWD database
  • Merlin Beaufils (2020): representation learning for web pages, with ASAPP (USA)

Supervision of research internships

++Master Level++

  • Stepan Svirin (2025-26): detecting errors in KGs using geometric information
  • Arpi Hunanyan (2025): studying the effect of emotions on the “Lost in the middle” effect for RAGs
  • Romain Harang (2020): language models for tweet geolocation (EGC 2021 workshop)

++Bachelor Level++

  • Andrei Stirbu (2026): transformer-based route reconstruction of AIS maritime tracks
  • Matheus Ferreira (2026): two-stages GNNs for node classification tasks
  • Eleonore Hasler (2024-25): geographic biases in LLMs
  • Jonas Treplin and Emre Belge (2023): unsupervised and semi-supervised hallucination detection in language models
  • Pénélope Forcioli and Liam Loughman (2023): detection of automatically generated text (ECNLP 2024)
  • Ekaterina Borisova and Antonina Mijatovic (2022): visual word sense disambiguation with multimodal models (SemEval 2023)
  • Dzhal Antonov (2022): relation classification in scientific knowledge graphs (ESWC 2023 workshop)
  • Mark Daychman (2022): clustering CS-KG relations with neural embeddings

International Collaborations

I am Scientific Communications Coordinator and French Management Committee member of the COST Action KGELL, which gathers more than 200 members from over 35 countries. I collaborate regularly with:

  • Italy: Università di Cagliari (Diego Reforgiato Recupero), Università di Genova (Barbara Catania, Giovanna Guerrini)
  • Spain: Universidad Politécnica de Valencia (Emilio Sanchis, Paolo Rosso), Universidad de Sevilla (Daniel Ayala, Agustín Borrego), Universidad de Málaga (María del Mar Roldán)
  • United Kingdom: Open University, Knowledge Media Institute (Francesco Osborne)
  • Mexico: UAM (Belém Priego Sánchez), INAOE Puebla (Manuel Montes, Luis Villaseñor)
  • Ireland: TU Dublin (Fernando Pérez-Téllez)
  • Germany: Coburg University of Applied Sciences (Jochen Leidner)

Visiting professors I have hosted at LIPN: Francesco Osborne (2017), Manuel Montes (2018), Fernando Pérez-Téllez (2019), Ana Maria Ghiran (Babeș-Bolyai University, 2023), Belém Priego (2024).

Scientific responsibilities

  • Member of the Scientific Council of LabEx EFL (InIdEx since 2025), since 2013
  • Elected member of the LIPN laboratory council (2016; since September 2024)
  • Member, and for a period chair, of the LIPN International Relations Commission (2017-2024)
  • Member of the Steering Committee for Human Language Technologies of the AFIA (2019-2024)
  • Member of the Steering Committee for SemEval-2020 and SemEval-2021
  • Member of the mentor board for the student mentorship program at ECIR 2016
  • Member of the jury for the ATALA best PhD thesis prize, 2019
  • Guest evaluator for the Frontier Science 2019 Call (Conacyt, Mexico)
  • Faculty selection committees: LIPN (2023, 2024), LIX “Monge” position (2025), IUT de Metz (2025), IUT de Montreuil (2025)

PhD juries

  • Fernando Luis Sola Espinosa, Universidad de Sevilla, 2026 (reviewer)
  • Abdelhak Kelious, Université de Lorraine, 2025 (reviewer)
  • Samuel Guilluy, Université de Rennes, 2024 (examiner)
  • Chuanming Dong, IGN, 2023 (examiner)
  • Danilo Dessì, Università di Cagliari, 2020 (examiner)
  • Ludovic Moncla, Université de Pau et des Pays de l’Adour, 2015 (examiner)
  • Julián Moreno Schneider, Universidad Carlos III de Madrid, 2015 (reviewer)

Reviewing

Education

  • HDR (Habilitation à Diriger les Recherches), Université Sorbonne Paris Nord, 18 March 2025. Improving access to Scientific Literature with Knowledge Graphs and Large Language Models. Jury: Nathalie Aussenac-Gilles, Thierry Charnois, Matthieu Constant (reviewer), Roberto Navigli (reviewer), Nathalie Pernelle, Mathieu Roche (reviewer), Fatiha Saïs.
  • PhD “cum laude” in Pattern Recognition and Artificial Intelligence, Universidad Politécnica de Valencia, 29 October 2010, European doctor mention. Toponym Disambiguation in Information Retrieval, supervised by Paolo Rosso. Jury: Paul Clough, Ross Purves, Emilio Sanchis, Mark Sanderson, Diana Santos. RiuNet
  • Laurea (B.Sc. + M.Sc.) in Computer Science, Università degli Studi di Genova, 2004. Knowledge-based Word Sense Disambiguation, supervised by Giorgio Delzanno.

Awards and grants

  • Best paper mention at EGC2024
  • Best poster award at the 5th Franco-Polish forum for Research and Innovation, 2019
  • Visiting professor grant from the University of Cagliari, May 2017
  • GVA grant for a 3-month doctoral stay at FBK-IRST, Trento, under the supervision of Bernardo Magnini, 2009
  • FPI grant from Generalitat Valenciana for the PhD program at Universidad Politécnica de Valencia, 2006-2010
  • DR78 grant from Regione Liguria for the completion of the B.Sc. thesis in a foreign research center, 2002

CV

A CV in French (last update: March 2026).