CoChat
Sign Up
Cited by 137Natural Language Processing TechniquesTopic ModelingSyntax, Semantics, Linguistic Variation

The DOI matches its Crossref record; 61 of the 66 references it cites were found in a registry or an index, and 1 are not checked yet.

A Systematic Assessment of Syntactic Generalization in Neural Language Models

Open access
Hu, Jennifer, Gauthier, Jon, Qian, Peng, Wilcox, Ethan, Levy, Roger·2020·Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
DOI 10.18653/v1/2020.acl-main.158
Open this paper in CoChat — free accountOpen PDF

Abstract

While state-of-the-art neural network models continue to achieve lower perplexity scores on language modeling benchmarks, it remains unknown whether optimizing for broad-coverage predictive performance leads to human-like syntactic knowledge.Furthermore, existing work has not provided a clear picture about the model properties required to produce proper syntactic generalizations.We present a systematic evaluation of the syntactic knowledge of neural language models, testing 20 combinations of model types and data sizes on a set of 34 English-language syntactic test suites.We find substantial differences in syntactic generalization performance by model architecture, with sequential models underperforming other architectures.Factorially manipulating model architecture and training dataset size (1M-40M words), we find that variability in syntactic generalization performance is substantially greater by architecture than by dataset size for the corpora tested in our experiments.Our results also reveal a dissociation between perplexity and syntactic generalization performance.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.18653/v1/2020.acl-main.158

Cites

61 of 62 references verified0 citation errors · 0 not checkable

Statistics

Figures

Image 1
CC-BY · From the PDF
Figure 1 Average SG score by model class. Asterisks denote off-the-shelf models. Error bars denote bootstrapped 95% confidence intervals of the mean.
Figure 1 Average SG score by model class. Asterisks denote off-the-shelf models. Error bars denote bootstrapped 95% confidence intervals of the mean.CC-BY · From the PDF
Figure 2 Relationship between SG score and perplexity on our held-out BLLIP test set for each model.
Figure 2 Relationship between SG score and perplexity on our held-out BLLIP test set for each model.CC-BY · From the PDF
Figure 3 Main results of our controlled evaluation of model class and dataset size. SG score varies more by model class (left) than by training dataset size (right).
Figure 3 Main results of our controlled evaluation of model class and dataset size. SG score varies more by model class (left) than by training dataset size (right).CC-BY · From the PDF
Figure 4 Controlled evaluation results, split across test suite circuits. Circuit-level differences in SG score vary more by model class (left) than by training dataset size (right).
Figure 4 Controlled evaluation results, split across test suite circuits. Circuit-level differences in SG score vary more by model class (left) than by training dataset size (right).CC-BY · From the PDF
Figure 5 Evaluation results on all models, split across test suite circuits.
Figure 5 Evaluation results on all models, split across test suite circuits.CC-BY · From the PDF
Figure 6 SG score on the pairs of test suites with and without intervening modifiers: Center Embedding, Cleft, MVRR, NPZ-Ambiguous, and NPZ-Object.
Figure 6 SG score on the pairs of test suites with and without intervening modifiers: Center Embedding, Cleft, MVRR, NPZ-Ambiguous, and NPZ-Object.CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

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

Show all 66 references

Research works printed but not linked: 6

  1. Andrew Carnie. 2012. Syntax: A generative introduc- tion, volume 18. John Wiley & Sons.
  2. Anastasia Giannakidou. 2011. Negative and positive polarity items: Variation, licensing, and composi- tionality. In Semantics: An international hand- book of natural language meaning, volume 3, pages 1660-1712. Berlin: Mouton de Gruyter.
  3. Mario Giulianelli, Jack Harding, Florian Mohnert, Dieuwke Hupkes, and Willem Zuidema. 2018. Un- der the hood: Using diagnostic classifiers to in- vestigate and improve how language models track agreement information. In Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and In- terpreting Neural Networks for NLP, pages 240- 248.
  4. Tanya Reinhart. 1981. Definite NP anaphora and c- command domains. Linguistic Inquiry, 12(4):605- 635.
  5. Ethan Wilcox, Jon Gauthier, Jennifer Hu, Peng Qian, and Roger P. Levy. 2020. Evaluating neural net- works as models of human online language process- ing. In Proceedings of the 42nd Meeting of the Cog- nitive Science Society (CogSci 2020). To appear.
  6. Ethan Wilcox, Peng Qian, Richard Futrell, Miguel Ballestros, and Roger P. Levy. 2019b. Structural su- pervision improves learning of non-local grammati- cal dependencies. In Proceedings of the 2019 Con- ference of the North American Chapter of the Asso- ciation for Computational Linguistics: Human Lan- guage Technologies, pages 3302-3312, Minneapo- lis, Minnesota.

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Hu, J., Gauthier, J., Qian, P., Wilcox, E., & Levy, R.
Affiliations
  • Massachusetts Institute of Technology (US)
  • Harvard University (US)
ORCID
Venue
Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
Publisher
Association for Computational Linguistics
Published
2020
Pages
1725-1744
ISSN
0736-587X (ISSN-L)
DOI
10.18653/v1/2020.acl-main.158
OpenAlex
W3034510440
Cited by
134
References
66
Citations by year
2026: 11 · 2025: 10 · 2024: 17 · 2023: 40 · 2022: 26 · 2021: 25 · 2020: 5
FWCI
9.17
Type
conference paper
Language
en
Open access
Licence
cc-by
Locations
Proceedings of the conference - Association for Computational Linguistics. Meeting (publishedVersion, cc-by)
Funders
  • National Science Foundation
  • Open Philanthropy Project
  • National Institutes of Health
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Social Sciences › Arts and Humanities › Language and Linguistics › Syntax, Semantics, Linguistic Variation
Keywords
syntactic generalization, neural language models, syntactic knowledge, model architecture, training data size, perplexity, English syntax, sequential model, language model evaluation

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