Projects

MapAIE/CHAI Since 2023	Mapping and exploring ethics of AI
	Tiphaine Viard, Mélanie Gornet, Simon Delarue, Maria Boritchev, Matthieu Labeau, Aina Garì Soler
	We collect and analyze charters, documents and manifestos about "ethical AI", and use them as a defining corpus for AI ethics topics of interest. More information here: https://mapaie.telecom-paris.fr/. Relevant publications: Mapping AI ethics: a meso-scale analysis of its charters and manifestos, Gornet, Mélanie and Delarue, Simon and Boritchev, Maria and Viard, Tiphaine. ACM Conference on Fairness, Accountability and Transparency Cartographier l'intelligence artificielle : collecte et analyse d'un corpus de chartes et manifestes sur l'éthique de l'IA, Viard, Tiphaine and Delarue, Simon and Gornet, Mélanie and Boritchev, Maria. Conférence Nationale Intelligence Artificielle (CNIA)

ANAGRAM Since 2024	AMR annotation of the DinG corpus
	Jeongwoo Kang, Maria Boritchev, Maximin Coavoux
	We are annotating a semantic corpus in French. We annotate the DinG corpus of transcriptions of spontaneous dialogues between players of Catan, in Abstract Meaning Representation (AMR), one of the most popular meaning representation frameworks. Since AMR does not cover some features of dialogue dynamics, we extend the framework to better represent spoken language as well as sentence structures specific to French. In addition, we provide an annotation guide detailing these extensions. Finally, we publish our corpus under a Creative Commons license (CC-SA-BY). Our work contributes to the development of semantic resources for French dialogue. We collected and transcribed 10 recordings of people playing the boardgame Catan, in French. We share the transcribed and anonymised data. We use this data as support for studying questions and answers in spontaneous French oral multilogues. Relevant publications: To be presented at TALN 2025: ding-01 :ARG0 -- un corpus AMR pour le français parlé spontané, Jeongwoo Kang, Maria Boritchev and Maximin Coavoux.

DinG Since 2019	Dialogues in Games corpus
	Maria Boritchev, Maxime Amblard
	We collected and transcribed 10 recordings of people playing the boardgame Catan, in French. We share the transcribed and anonymised data. We use this data as support for studying questions and answers in spontaneous French oral multilogues. More information here and here: https://team.inria.fr/semagramme/projects/corpus-ding/ Relevant publications: A Multi-Party Dialogue Ressource in French, Boritchev, Maria and Amblard, Maxime. Proceedings of the Thirteenth Language Resources and Evaluation Conference

AMR w/ Orange Since 2023	Error analysis, questions, and AMR
	Maria Boritchev, Johannes Heinecke, Frédéric Herlédan
	We conduct an error analysis of AMR parsers' outputs in order to characterise these errors and come up with pre- and post- processing heuristics to patch these. We work on data in English, French, German, Spanish, Italian, and Polish. We analyse how adding few sentence-type specific annotations can steer the model to improve parsing in the case of questions in English. Relevant publications: Question-parsing with Abstract Meaning Representation enhanced by adding small datasets, Heinecke, Johannes and Boritchev, Maria and Herledan, Frédéric. Proceedings of the Joint 25th Nordic Conference on Computational Linguistics and 11th Baltic Conference on Human Language Technologies (NoDaLiDa/Baltic-HLT 2025) Error Exploration for Automatic Abstract Meaning Representation Parsing, Boritchev, Maria and Heinecke, Johannes. 15th International Conference on Computational Semantics