Using Computational Models to Test Syntactic Learnability
<jats:title>Abstract</jats:title> <jats:p>We study the learnability of English filler–gap dependencies and the “island” constraints on them by assessing the generalizations made by autoregressive (incremental) language models that use deep learning to predict the ne...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
MIT Press
2023
|
| Online Access: | https://hdl.handle.net/1721.1/150009 |
| _version_ | 1826204628528660480 |
|---|---|
| author | Wilcox, Ethan Gotlieb Futrell, Richard Levy, Roger |
| author2 | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences |
| author_facet | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Wilcox, Ethan Gotlieb Futrell, Richard Levy, Roger |
| author_sort | Wilcox, Ethan Gotlieb |
| collection | MIT |
| description | <jats:title>Abstract</jats:title>
<jats:p>We study the learnability of English filler–gap dependencies and the “island” constraints on them by assessing the generalizations made by autoregressive (incremental) language models that use deep learning to predict the next word given preceding context. Using factorial tests inspired by experimental psycholinguistics, we find that models acquire not only the basic contingency between fillers and gaps, but also the unboundedness and hierarchical constraints implicated in the dependency. We evaluate a model’s acquisition of island constraints by demonstrating that its expectation for a filler–gap contingency is attenuated within an island environment. Our results provide empirical evidence against the Argument from the Poverty of the Stimulus for this particular structure.</jats:p> |
| first_indexed | 2024-09-23T12:58:28Z |
| format | Article |
| id | mit-1721.1/150009 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T12:58:28Z |
| publishDate | 2023 |
| publisher | MIT Press |
| record_format | dspace |
| spelling | mit-1721.1/1500092023-04-01T03:31:00Z Using Computational Models to Test Syntactic Learnability Wilcox, Ethan Gotlieb Futrell, Richard Levy, Roger Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences <jats:title>Abstract</jats:title> <jats:p>We study the learnability of English filler–gap dependencies and the “island” constraints on them by assessing the generalizations made by autoregressive (incremental) language models that use deep learning to predict the next word given preceding context. Using factorial tests inspired by experimental psycholinguistics, we find that models acquire not only the basic contingency between fillers and gaps, but also the unboundedness and hierarchical constraints implicated in the dependency. We evaluate a model’s acquisition of island constraints by demonstrating that its expectation for a filler–gap contingency is attenuated within an island environment. Our results provide empirical evidence against the Argument from the Poverty of the Stimulus for this particular structure.</jats:p> 2023-03-30T13:34:47Z 2023-03-30T13:34:47Z 2022 2023-03-30T13:32:15Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/150009 Wilcox, Ethan Gotlieb, Futrell, Richard and Levy, Roger. 2022. "Using Computational Models to Test Syntactic Learnability." Linguistic Inquiry. en 10.1162/LING_A_00491 Linguistic Inquiry Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf MIT Press MIT Press |
| spellingShingle | Wilcox, Ethan Gotlieb Futrell, Richard Levy, Roger Using Computational Models to Test Syntactic Learnability |
| title | Using Computational Models to Test Syntactic Learnability |
| title_full | Using Computational Models to Test Syntactic Learnability |
| title_fullStr | Using Computational Models to Test Syntactic Learnability |
| title_full_unstemmed | Using Computational Models to Test Syntactic Learnability |
| title_short | Using Computational Models to Test Syntactic Learnability |
| title_sort | using computational models to test syntactic learnability |
| url | https://hdl.handle.net/1721.1/150009 |
| work_keys_str_mv | AT wilcoxethangotlieb usingcomputationalmodelstotestsyntacticlearnability AT futrellrichard usingcomputationalmodelstotestsyntacticlearnability AT levyroger usingcomputationalmodelstotestsyntacticlearnability |