Job siteWelcome

Design and implement algorithms for event extraction (detection, classification, semantic role tagging) from French texts

Develop methods for event reconciliation (alignment, merging, conflict resolution) between multiple sources and prior knowledge bases

Contribute to the creation of annotated corpora and benchmarks for evaluating the proposed methods
Propose evaluation metrics tailored to the specific challenges of the tasks (e.g. handling partial events, information conflicts)
Contribute to the publication of results at international conferences (e.g. ACL, EMNLP, TALN) and in scientific journals

• Proficiency in programming languages: Python (PyTorch, TensorFlow, HuggingFace), NLP libraries (spaCy, Stanza, etc.), lower-level languages (Rust/C)
• Knowledge of techniques and ability to implement language models (LLMs): pre-training or fine-tuning, embeddings, transformers (BERT, RoBERTa, etc.)
• Machine learning: supervised methods, graph neural networks, data augmentation
• Computational linguistics: syntactic/semantic parsing (UD, AMR), coreference resolution
• Experience with knowledge graphs or symbolic systems
• Proficiency in collaborative annotation tools (e.g. INCEpTION, Prodigy)
• Interest in the ethical issues surrounding disinformation or the analysis of sensitive data
• Independence and scientific rigour: the ability to propose innovative solutions and validate them experimentally
• A collaborative spirit: the ability to work effectively within a multidisciplinary team (linguists, engineers, researchers)
• Intellectual curiosity: an interest in open scientific challenges (e.g. semantic alignment, evaluation of large language models)
• Writing skills: the ability to draft technical reports and contribute to research articles
• A minimum of a Master’s degree is required, with experience in AI/NLP
  A PhD in Natural Language Processing (NLP), Computer Science or a related field would be an advantage
• Experience in at least one of the following areas: modern NLP (large language models (LLMs), embeddings, transformers), machine learning (supervised methods, graph neural networks, data augmentation), computational linguistics (syntactic/semantic parsing (UD, AMR), coreference resolution)