Marine Carpuat

Professor · University of Maryland
Department of Computer Science & UMIACS

Director, CLIP Lab  ·  Member, TRAILS & AIM

Marine Carpuat
Human-Centered AI NLP LLMs Multilinguality Translation Speech & Multimodality

My goal: build AI tools that help people communicate and act in the world, no matter what language they speak.

This leads to the question: does AI-mediated communication actually help people understand each other, or does it only appear to? Across settings, we observe a gap between surface-level performance and real communicative success. Understanding and addressing this gap is the core focus of my research.

My group approaches this problem from multiple angles: from the internal representations that enable or undermine multilingual capabilities in large language models, to how real people perceive, rely on, and are sometimes misled by AI-mediated communication. We develop new techniques and evaluation methods, and complement this technical work with human-centered and interdisciplinary studies.

Right now, my group is most excited about three directions: designing evaluation frameworks for AI-mediated communication that connect empirical findings to technology development; extending this work to conversational settings, where communication is spoken and interactive; and using models' internal signals to improve both model capabilities and how people rely on model outputs.

Recent work spanning user perception and reliance on machine translation, quality signals for text and speech, and generalization in multilingual LLMs. Full list on Google Scholar →

Framing paper

An interdisciplinary approach to human-centered machine translation

Carpuat, Asscher, Bali, Bentivogli, Blain, Bowker, et al.  ·  EMNLP 2025

User perception and reliance
Quality estimation for AI translation and speech models
Generalization and alignment in multilingual LLMs

Marine Carpuat is a Professor in the Department of Computer Science at the University of Maryland, where she is also a member of UMIACS and director of the CLIP Lab. Her research focuses on developing AI techniques that help people communicate across languages, and studying whether those systems actually succeed. Her work spans foundational NLP methods, evaluation methodology, and human-centered studies of how people perceive and rely on AI-generated translations. She has published extensively at venues including ACL, EMNLP, NAACL, and CHI, and has received paper awards from *SEM, TALN, and EMNLP. She served as Program Co-Chair of NAACL 2022. Before joining UMD, she was a Research Officer at the National Research Council Canada. She received her PhD from the Hong Kong University of Science and Technology and a diplôme d'ingénieur from the French grande école Supélec.

Office
Iribe Center 4130, University of Maryland, College Park
Affiliations
Scholar