Seminar@SystemX animé par Zied Bouraoui

Zied Bouraoui (University of Artois – CRIL, CNRS) animera un Seminar@SystemX sur le thème « Learning Conceptual Representations from Language Models », le 20 mars 2025 de 14h à 15h.

>> Lien de connexion au webinaire <<

Résumé

Capturing the semantic meaning of concepts in static vector representations remains a key challenge in applications that require word meaning to be modeled without contextual information. Static embeddings are essential for tasks such as zero-shot image classification, zero-shot entity typing, ontology alignment and completion, taxonomy learning, and query representation in information retrieval systems. In this talk, I will explore strategies for learning effective concept and relation representations from language models and provide an overview of how these embeddings can enhance downstream applications.

Biographie

Zied Bouraoui is a professor of Computer Science at the University of Artois and a researcher at CRIL (CNRS UMR 8188). His primary research interests include commonsense reasoning, reasoning with imperfect information, and natural language processing. He has co-authored over 40 peer-reviewed papers published in top-tier AI journals and conferences (e.g., AAAI, IJCAI, KRR, ACL, SIGIR, EMNLP), many of which have received awards at prestigious conferences. His research is supported by several projects, including the Chair IA Be4musIA, ANR JCJC ERIANA (2023-2027), ANR VIVAH (2020-2025), and the European H2020 ICT-48 Trustworthy AI program.

 

Inscription

• Pour être informé de nos prochains Seminar@SystemX, rendez-vous ici.