PhD & Assoc. Prof., Inalco, Paris, France

"The question whether machines can think is about as relevant as the question whether submarines can swim" (E. Dijkstra)

Research Interests & Expertise

As an Associate Professor at Inalco and AI Officer for the institution , my research focuses on bridging the gap between linguistic data and computational models. My current work centers on:

• Large Language Models (LLMs) & Generative AI: Monitoring capabilities and implementing policies for research and education
• Multilingualism & Low-Resource Languages: Developing NLP tools for languages such as Armenian, Bambara, and Quechua
• Information Extraction: Specializing in Named Entity Recognition (NER) and Linking
• Text Mining & Discourse Analysis: Leveraging statistical methods for large-scale linguistic data processing

PhD Research

My doctoral research explored the synergy between knowledge-based and data-driven approaches for Named Entity Recognition, specifically within oral transcripts . I developed the mXS system, which utilizes an exhaustive pattern mining approach to discover hierarchical sequential patterns in annotated corpora.

The originality of this method lies in its "local instruction" strategy: rather than categorizing entire word sequences, the system induces transducers (annotation rules) that precisely mark the boundaries (beginning and end tags) of entities . This approach offers a robust alternative to traditional word-level classification. You can find more details and the French implementation on the mXS project page.

If you are exploring the field of Named Entity Recognition, I recommend the following foundational readings:

• Grishman & Sundheim (1996): Message Understanding Conference-6: A Brief History.
• Nadeau & Sekine (2007): A survey of named entity recognition and classification
• Bikel et al. (1999): Learning algorithms for NER
• Tjong Kim Sang & De Meulder (2003): The CoNLL-2003 shared task (Language-independent NER)
• McCallum & Li (2003): Early results using Conditional Random Fields (CRFs)
• Ratinov & Roth (2009): Design challenges and misconceptions in NER