Emergent linguistic structure in artificial neural networks trained by self-supervision
Open accessThe DOI matches its Crossref record; 64 of the 64 references it cites were found in a registry or an index.
Abstract
This paper explores the knowledge of linguistic structure learned by large artificial neural networks, trained via self-supervision, whereby the model simply tries to predict a masked word in a given context. Human language communication is via sequences of words, but language understanding requires constructing rich hierarchical structures that are never observed explicitly. The mechanisms for this have been a prime mystery of human language acquisition, while engineering work has mainly proceeded by supervised learning on treebanks of sentences hand labeled for this latent structure. However, we demonstrate that modern deep contextual language models learn major aspects of this structure, without any explicit supervision. We develop methods for identifying linguistic hierarchical structure emergent in artificial neural networks and demonstrate that components in these models focus on syntactic grammatical relationships and anaphoric coreference. Indeed, we show that a linear transformation of learned embeddings in these models captures parse tree distances to a surprising degree, allowing approximate reconstruction of the sentence tree structures normally assumed by linguists. These results help explain why these models have brought such large improvements across many language-understanding tasks.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
Glove: Global Vectors for Word Representation
Pennington, J., Socher, R., & Manning, C. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1162 · Cited by 34,071
Self-citationDistributed Representations of Words and Phrases and their Compositionality
Mikolov, T., Sutskever, I., Chen, K., Corrado, G., & Dean, J. (2013) · arXiv · doi:10.48550/arxiv.1310.4546 · Cited by 24,533
Neural Machine Translation by Jointly Learning to Align and Translate
Bahdanau, D., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1409.0473 · Cited by 16,213
Advances in Neural Information Processing Systems 19
(2007) · The MIT Press eBooks · doi:10.52202/085713 · Cited by 15,539
Advances in Neural Information Processing Systems 29
Teymur, O., Zygalakis, K., & Calderhead, B. (2016) · Cited by 13,887
Advances in Neural Information Processing Systems 25
Zhang, Y., Sutton, C. A., Storkey, A., & Zoubin Ghahramani, A. (2012) · Cited by 12,653
Deep Contextualized Word Representations
Peters, M., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers) · doi:10.18653/v1/n18-1202 · Cited by 11,983
Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1412.3555 · Cited by 11,008
Cited by
A primer in BERTology: What we know about how BERT works
Rogers, A., Kovaleva, O., & Rumshisky, A. (2020) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00349 · Cited by 1,552
Pre-trained models: Past, present and future
Han, X., Zhang, Z., Ding, N., Gu, Y., Liu, X., Huo, Y., Qiu, J., Yao, Y., Zhang, A., Zhang, L., et al. (2021) · AI Open · doi:10.1016/j.aiopen.2021.08.002 · Cited by 994
Shared computational principles for language processing in humans and deep language models
Goldstein, A., Zada, Z., Buchnik, E., Schain, M., Price, A., Aubrey, B., Nastase, S. A., Feder, A., Emanuel, D., Cohen, A., et al. (2022) · Nature Neuroscience · doi:10.1038/s41593-022-01026-4 · Cited by 447
Self-citationA hierarchy of linguistic predictions during natural language comprehension
Heilbron, M., Armeni, K., Schoffelen, J.-M., Hagoort, P., & de Lange, F. P. (2022) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2201968119 · Cited by 343
More than a Feeling: Accuracy and Application of Sentiment Analysis
Hartmann, J., Heitmann, M., Siebert, C., & Schamp, C. (2023) · International Journal of Research in Marketing · doi:10.1016/j.ijresmar.2022.05.005 · Cited by 356
Dissociating language and thought in large language models
Mahowald, K., Ivanova, A. A., Blank, I. A., Kanwisher, N., Tenenbaum, J. B., & Fedorenko, E. (2024) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2024.01.011 · Cited by 345
Using large language models in psychology
Demszky, D., Yang, D., Yeager, D. S., Bryan, C. J., Clapper, M., Chandhok, S., Eichstaedt, J. C., Hecht, C., Jamieson, J., Johnson, M., et al. (2023) · Nature Reviews Psychology · doi:10.1038/s44159-023-00241-5 · Cited by 369
Brains and algorithms partially converge in natural language processing
Caucheteux, C., & King, J.-R. (2022) · Communications Biology · doi:10.1038/s42003-022-03036-1 · Cited by 318
“Further inspection of arti fi cial 8,68 and biological networks 10,28,69 remains necessary to further decompose them into interpretable features.”
BackgroundEvidence of a predictive coding hierarchy in the human brain listening to speech
Caucheteux, C., Gramfort, A., & King, J.-R. (2023) · Nature Human Behaviour · doi:10.1038/s41562-022-01516-2 · Cited by 308
Hartmann, J., Heitmann, M., Schamp, C., & Netzer, O. (2021) · Journal of Marketing Research · doi:10.1177/00222437211037258 · Cited by 184
Acquisition of chess knowledge in AlphaZero
McGrath, T., Kapishnikov, A., Tomašev, N., Pearce, A., Wattenberg, M., Hassabis, D., Kim, B., Paquet, U., & Kramnik, V. (2022) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2206625119 · Cited by 166
“Certain language models reproduce sentence parse trees (7).”
BackgroundSarstedt, M., Adler, S. J., Rau, L., & Schmitt, B. (2024) · Psychology & Marketing · doi:10.1002/mar.21982 · Cited by 149
“Furthermore, researchers should use probing to assess the output's robustness against, for example, linguistic features (Manning et al., 2020).”
BackgroundAn Interpretability Illusion for BERT
Bolukbasi, T., Pearce, A., Yuan, A., Coenen, A., Reif, E., Viégas, F., & Wattenberg, M. (2021) · arXiv · doi:10.48550/arxiv.2104.07143 · Cited by 113
What Artificial Neural Networks Can Tell Us about Human Language Acquisition
Warstadt, A., & Bowman, S. R. (2022) · Algebraic Structures in Natural Language · doi:10.1201/9781003205388-2 · Cited by 111
Do Large Language Models Know What Humans Know?
Trott, S., Jones, C., Chang, T., Michaelov, J., & Bergen, B. (2023) · Cognitive Science · doi:10.1111/cogs.13309 · Cited by 99
McCoy, R. T., Yao, S., Friedman, D., Hardy, M. D., & Griffiths, T. L. (2024) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2322420121 · Cited by 96
McClelland, J. L., Hill, F., Rudolph, M., Baldridge, J., & Schütze, H. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1910416117 · Cited by 91
Deep language algorithms predict semantic comprehension from brain activity
Caucheteux, C., Gramfort, A., & King, J.-R. (2022) · Scientific Reports · doi:10.1038/s41598-022-20460-9 · Cited by 91
Generating meaning: active inference and the scope and limits of passive AI
Pezzulo, G., Parr, T., Cisek, P., Clark, A., & Friston, K. (2024) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2023.10.002 · Cited by 80
“…predictive training – the latent variables of generative AI likewise come to re fl ect deep regularities (e.g., emergent linguistic structure in LLMs [42,43]) that might extend beyond the training domain (e.g., non-linguistic regularities for LLMs, such as the relations between looks and tastes).”
Goldstein, A., Grinstein-Dabush, A., Schain, M., Wang, H., Hong, Z., Aubrey, B., Nastase, S. A., Zada, Z., Ham, E., Feder, A., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-46631-y · Cited by 64
A hierarchy of linguistic predictions during natural language comprehension
(2020) · doi:10.1101/2020.12.03.410399 · Cited by 63
MoEfication: Transformer Feed-forward Layers are Mixtures of Experts
Zhang, Z., Lin, Y., Liu, Z., Li, P., Sun, M., & Zhou, J. (2022) · Findings of the Association for Computational Linguistics: ACL 2022 · doi:10.18653/v1/2022.findings-acl.71 · Cited by 48
Communicating artificial neural networks develop efficient color-naming systems
Chaabouni, R., Kharitonov, E., Dupoux, E., & Baroni, M. (2021) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2016569118 · Cited by 39
“16– 19) ,”
Natural Language Processing in Marketing
Hartmann, J., & Netzer, O. (2023) · Artificial Intelligence in Marketing · doi:10.1108/s1548-643520230000020011 · Cited by 38
Lampinen, A. (2024) · Computational Linguistics · doi:10.1162/coli_a_00525 · Cited by 32
“In this context, these experiments contribute to the growing evidence (e.g., Manning et al. 2020) that modern LMs can learn to process complex syntactic structures, even ones that challenge humans, with a relatively domain-general architecture and learning process applied to imperfect natural data.”
Lampinen, A. K. (2022) · arXiv · doi:10.48550/arxiv.2210.15303 · Cited by 32
van Duijn, M., van Dijk, B., Kouwenhoven, T., de Valk, W., Spruit, M., & vanderPutten, P. (2023) · Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL) · doi:10.18653/v1/2023.conll-1.25 · Cited by 24
Model-based analysis of brain activity reveals the hierarchy of language in 305 subjects
Caucheteux, C., Gramfort, A., & King, J.-R. (2021) · Findings of the Association for Computational Linguistics: EMNLP 2021 · doi:10.18653/v1/2021.findings-emnlp.308 · Cited by 21
The Life Cycle of Knowledge in Big Language Models: A Survey
Cao, B., Lin, H., Han, X., & Sun, L. (2023) · arXiv · doi:10.48550/arxiv.2303.07616 · Cited by 18
The Life Cycle of Knowledge in Big Language Models: A Survey
Cao, B., Lin, H., Han, X., & Sun, L. (2024) · Machine Intelligence Research · doi:10.1007/s11633-023-1416-x · Cited by 18
Generating Meaning: Active Inference and the Scope and Limits of Passive AI
Pezzulo, G., Parr, T., Cisek, P., Clark, A., & Friston, K. (2023) · doi:10.31234/osf.io/8xgzv · Cited by 8
Do Vision and Language Models Share Concepts? A Vector Space Alignment Study
Li, J., Kementchedjhieva, Y., Fierro, C., & Søgaard, A. (2024) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00698 · Cited by 6
“The literature on large-scale, pretrained models has reported seemingly emergent properties (Søgaard et al., 2018; Manning et al., 2020; Garneau et al., 2021; Teehan et al., 2022; Wei et al., 2022), many of which relate to induction of world knowledge.”
Background(2022) · doi:10.18653/v1/2022.conll-1.8 · Cited by 5
Hong, Z., Wang, H., Zada, Z., Gazula, H., Turner, D., Aubrey, B., Niekerken, L., Doyle, W., Devore, S., Dugan, P., et al. (2024) · doi:10.1101/2024.06.12.598513 · Cited by 1
Metadata
- Authors
- Manning, C. D., Clark, K., Hewitt, J., Khandelwal, U., & Levy, O.
- Venue
- Proceedings of the National Academy of Sciences
- Published
- 2020-06-03
- Volume
- 117
- Issue
- 48
- Pages
- 30046-30054
- ISSN
- 0027-8424 (ISSN-L)
- 1091-6490
- OpenAlex
- W3031914912
- Cited by
- 339
- References
- 64
- FWCI
- 18.37
- Type
- journal article
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.