John T. Hale
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
John T. Hale has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Adhiguna Kuncoro
University of Oxford
1 shared publication - Chris Dyer
Computer Science Division UC Berkeley London UK
1 shared publication - Jonathan R. Brennan
Department of Linguistics, University of Michigan, Ann Arbor, Michigan 48109
1 shared publication - Miloš Stanojević
Google DeepMind University of Georgia University of Georgia Jan Buys University of Cape Town University of Georgia
1 shared publication
Latest Publications
- Open Access
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 17106-17119
Do LLMs learn a true syntactic universal?
John T. Hale, Miloš Stanojević. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 17106-17119
Open AccessDo LLMs learn a true syntactic universal?
John T. Hale, Miloš Stanojević. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.
- Open Access
Finding Syntax in Human Encephalography with Beam Search
Recurrent neural network grammars (RNNGs) are generative models of (tree,string) pairs that rely on neural networks to evaluate derivational choices. Parsing with them using beam search yields a variety of incremental complexity metrics such as word surprisal and parser action count. When used as regressors against human electrophysiological responses to naturalistic text, they derive two amplitude effects: an early peak and a P600-like later peak. By contrast, a non-syntactic neural language model yields no reliable effects. Model comparisons attribute the early peak to syntactic composition within the RNNG. This pattern of results recommends the RNNG+beam search combination as a mechanistic model of the syntactic processing that occurs during normal human language comprehension.
Open AccessFinding Syntax in Human Encephalography with Beam Search
Recurrent neural network grammars (RNNGs) are generative models of (tree,string) pairs that rely on neural networks to evaluate derivational choices. Parsing with them using beam search yields a variety of incremental complexity metrics such as word surprisal and parser action count. When used as regressors against human electrophysiological responses to naturalistic text, they derive two amplitude effects: an early peak and a P600-like later peak. By contrast, a non-syntactic neural language model yields no reliable effects. Model comparisons attribute the early peak to syntactic composition within the RNNG. This pattern of results recommends the RNNG+beam search combination as a mechanistic model of the syntactic processing that occurs during normal human language comprehension.