Timothy O'Donnell
Also known as Timothy J. O'Donnell
Timothy O'Donnell is a computational linguist whose AI research studies how people learn and process language. He earned a PhD from Harvard, developed Fragment Grammars for language productivity, and presented that work in his 2015 book, Productivity and Reuse in Language: A Theory of Linguistic Computation and Storage. At McGill, he was an investigator on a project about robust unsupervised language learning and coauthored a 2022 system for building interpretable models of human language. His 2025 papers examined causal relations among word patterns and ways to convert token-level language models into character-level models. As of October 2026, he is an associate professor of linguistics at McGill, a Canada CIFAR AI Chair at Mila.
- Now at
- McGill University and Mila
- Works
- 139
- Citations
- 1,823
- h-index
- 18
Publication numbers from OpenAlex, as of 27 October 2026.
Timeline
-
2011
Fragment Grammars
In 2011, his Harvard psychology dissertation develops Fragment Grammars, a Bayesian model of language productivity and reuse. He holds a PhD from Harvard University.
-
2015
Productivity and Reuse in Language
In 2015, he publishes Productivity and Reuse in Language: A Theory of Linguistic Computation and Storage, presenting that work in book form.
-
2018–2022
Unsupervised language learning
From 2018 to 2022, he is the investigator for McGill’s NSERC Discovery project Towards Robust Unsupervised Language Learning. The project aims to build systems that generalize in human-like ways through theoretical linguistics and probabilistic models.
-
2019
A Canada CIFAR AI Chair
In 2019, he is named among six new Canada CIFAR AI Chairs affiliated with McGill’s School of Computer Science.
-
2022
A Bayesian language system
In 2022, he coauthors Synthesizing theories of human language with Bayesian program induction. The framework combines Bayesian inference and program synthesis to build human-interpretable models across 70 datasets from 58 languages.
-
2025
Causal discovery across languages
In 2025, he coauthors Unzipping the Causality of Zipf’s Law and Other Lexical Trade-offs. The paper applies causal discovery to word length, frequency, phonotactic complexity, and morphological irregularity across 25 languages.
-
2025
Character-level language models
In 2025, he coauthors From Language Models over Tokens to Language Models over Characters, which presents exact and approximate algorithms for converting token-level models into character-level models.
-
As of October 2026
McGill associate professor
As of October 2026, he is an associate professor in McGill’s Department of Linguistics, a Canada CIFAR AI Chair at Mila. His expertise covers computational models of language learning and processing, mathematical linguistics, and probabilistic inference.
Career
- NowMcGill UniversityAssociate Professor
- NowMilaCanada CIFAR AI Chair
Education
- 2011Harvard UniversityPhD
Sources and credits
Photo: We have no published photo of Timothy O'Donnell yet. Send us a free-licensed photo.
- Productivity and Reuse in Language - McGill NLP
- Timothy O'Donnell | Linguistics - McGill University
- CiNii 図書 - Productivity and reuse in language : a theory of linguistic computation and storage
- Projects & Grants | Linguistics - McGill University
- Announcement of New Canada CIFAR AI Chairs
- Synthesizing theories of human language with Bayesian program induction | Nature Communications
- Unzipping the Causality of Zipf’s Law and Other Lexical Trade-offs - ACL Anthology
- From Language Models over Tokens to Language Models over Characters - Proceedings of Machine Learning Research
- People | Linguistics - McGill University