Congratulations to Nicolas Hubert, doctoral student in Loria’s BIRD team and ERPI laboratory, who won two Best Paper Awards at ESWC 2024 (Extended Semantic Web Conference), the leading international conference in the field of the semantic web!
The paper proposes to enrich loss functions for link prediction in knowledge graphs, taking into account the semantic validity of negative triples. In particular, Nicolas Hubert shows that negative triples that respect relation signatures (domain and co-domain) are high-quality negatives, and that they contribute experimentally to improving the semantic and predictive capacities of the models analyzed.