Léo Jacqmin

I am a PhD candidate at Orange and Aix-Marseille University working on conversational AI. I am happy to be working with Lina M. Rojas Barahona (Orange) and Benoit Favre (AMU). My thesis focuses on dialogue state tracking, a task concerned with capturing the user’s needs throughout the course of a dialogue. I am especially interested in exploring novel state representations that can suit the practical requirements of task-oriented dialogue systems.

Previously, I received my master’s degree in Natural Language Processing from the University of Lorraine. I interned at Orange where I worked on multilingual aspect-based sentiment analysis as part of my master’s research project.

In my free time I enjoy surfing, reading, meditating, and much more.

News

Jan 17, 2022	Attending ALPS 2022, an NLP winter school.
Oct 1, 2021	Started my PhD journey.
Sep 10, 2021	One paper accepted at W-NUT 2021 co-located with EMNLP 2021.

Selected Publications

SpanAlign: Efficient Sequence Tagging Annotation Projection into Translated Data applied to Cross-Lingual Opinion Mining

Jacqmin, Léo, Marzinotto, Gabriel, Gromada, Justyna, Szczekocka, Ewelina, Kołodyński, Robert, and Damnati, Géraldine

In Proceedings of the Seventh Workshop on Noisy User-generated Text (W-NUT 2021) Nov 2021

Abs PDF

Following the increasing performance of neural machine translation systems, the paradigm of using automatically translated data for cross-lingual adaptation is now studied in several applicative domains. The capacity to accurately project annotations remains however an issue for sequence tagging tasks where annotation must be projected with correct spans. Additionally, when the task implies noisy user-generated text, the quality of translation and annotation projection can be affected. In this paper we propose to tackle multilingual sequence tagging with a new span alignment method and apply it to opinion target extraction from customer reviews. We show that provided suitable heuristics, translated data with automatic span-level annotation projection can yield improvements both for cross-lingual adaptation compared to zero-shot transfer, and data augmentation compared to a multilingual baseline.