Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages

Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages

Despite the fact that Transformers perform well in NLP tasks, recent studies suggest that self-attention is theoretically limited in learning even some regular and context-free languages. These findings motivated us to think about their implications in modeling natural language, which is hypothesize...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2023-10
Hauptverfasser: Wang, Shunjie, Steinert-Threlkeld, Shane
Format: Artikel
Sprache:eng
Schlagworte:
Context free languages
Transformers
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Wang, Shunjie
Steinert-Threlkeld, Shane
description Despite the fact that Transformers perform well in NLP tasks, recent studies suggest that self-attention is theoretically limited in learning even some regular and context-free languages. These findings motivated us to think about their implications in modeling natural language, which is hypothesized to be mildly context-sensitive. We test the Transformer's ability to learn mildly context-sensitive languages of varying complexities, and find that they generalize well to unseen in-distribution data, but their ability to extrapolate to longer strings is worse than that of LSTMs. Our analyses show that the learned self-attention patterns and representations modeled dependency relations and demonstrated counting behavior, which may have helped the models solve the languages.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2861510015</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2861510015</sourcerecordid><originalsourceid>FETCH-proquest_journals_28615100153</originalsourceid><addsrcrecordid>eNqNyrEKwjAUQNEgCBbtPwQcnAppYmpXKRWHOtm9RHwtKTHRvKTYv9fBD3C6w7kLknAh8qzcc74iKeLIGOPFgUspEtLUkzJRBW0H2nplsXf-AX6H9HjTRoeZBkcbUN7SizZ3M9PK2QDvkF3Bog56AtooO0Q1AG7IslcGIf11Tbanuq3O2dO7VwQM3eiit1_qeFnkMmcsl-K_6wNnJz2Z</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2861510015</pqid></control><display><type>article</type><title>Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages</title><source>Free E- Journals</source><creator>Wang, Shunjie ; Steinert-Threlkeld, Shane</creator><creatorcontrib>Wang, Shunjie ; Steinert-Threlkeld, Shane</creatorcontrib><description>Despite the fact that Transformers perform well in NLP tasks, recent studies suggest that self-attention is theoretically limited in learning even some regular and context-free languages. These findings motivated us to think about their implications in modeling natural language, which is hypothesized to be mildly context-sensitive. We test the Transformer's ability to learn mildly context-sensitive languages of varying complexities, and find that they generalize well to unseen in-distribution data, but their ability to extrapolate to longer strings is worse than that of LSTMs. Our analyses show that the learned self-attention patterns and representations modeled dependency relations and demonstrated counting behavior, which may have helped the models solve the languages.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Context free languages ; Transformers</subject><ispartof>arXiv.org, 2023-10</ispartof><rights>2023. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>780,784</link.rule.ids></links><search><creatorcontrib>Wang, Shunjie</creatorcontrib><creatorcontrib>Steinert-Threlkeld, Shane</creatorcontrib><title>Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages</title><title>arXiv.org</title><description>Despite the fact that Transformers perform well in NLP tasks, recent studies suggest that self-attention is theoretically limited in learning even some regular and context-free languages. These findings motivated us to think about their implications in modeling natural language, which is hypothesized to be mildly context-sensitive. We test the Transformer's ability to learn mildly context-sensitive languages of varying complexities, and find that they generalize well to unseen in-distribution data, but their ability to extrapolate to longer strings is worse than that of LSTMs. Our analyses show that the learned self-attention patterns and representations modeled dependency relations and demonstrated counting behavior, which may have helped the models solve the languages.</description><subject>Context free languages</subject><subject>Transformers</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNyrEKwjAUQNEgCBbtPwQcnAppYmpXKRWHOtm9RHwtKTHRvKTYv9fBD3C6w7kLknAh8qzcc74iKeLIGOPFgUspEtLUkzJRBW0H2nplsXf-AX6H9HjTRoeZBkcbUN7SizZ3M9PK2QDvkF3Bog56AtooO0Q1AG7IslcGIf11Tbanuq3O2dO7VwQM3eiit1_qeFnkMmcsl-K_6wNnJz2Z</recordid><startdate>20231019</startdate><enddate>20231019</enddate><creator>Wang, Shunjie</creator><creator>Steinert-Threlkeld, Shane</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20231019</creationdate><title>Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages</title><author>Wang, Shunjie ; Steinert-Threlkeld, Shane</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_28615100153</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Context free languages</topic><topic>Transformers</topic><toplevel>online_resources</toplevel><creatorcontrib>Wang, Shunjie</creatorcontrib><creatorcontrib>Steinert-Threlkeld, Shane</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Wang, Shunjie</au><au>Steinert-Threlkeld, Shane</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages</atitle><jtitle>arXiv.org</jtitle><date>2023-10-19</date><risdate>2023</risdate><eissn>2331-8422</eissn><abstract>Despite the fact that Transformers perform well in NLP tasks, recent studies suggest that self-attention is theoretically limited in learning even some regular and context-free languages. These findings motivated us to think about their implications in modeling natural language, which is hypothesized to be mildly context-sensitive. We test the Transformer's ability to learn mildly context-sensitive languages of varying complexities, and find that they generalize well to unseen in-distribution data, but their ability to extrapolate to longer strings is worse than that of LSTMs. Our analyses show that the learned self-attention patterns and representations modeled dependency relations and demonstrated counting behavior, which may have helped the models solve the languages.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2023-10
issn 2331-8422
language eng
recordid cdi_proquest_journals_2861510015
source Free E- Journals
subjects Context free languages
Transformers
title Evaluating Transformer's Ability to Learn Mildly Context-Sensitive Languages
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-20T07%3A33%3A16IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Evaluating%20Transformer's%20Ability%20to%20Learn%20Mildly%20Context-Sensitive%20Languages&rft.jtitle=arXiv.org&rft.au=Wang,%20Shunjie&rft.date=2023-10-19&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2861510015%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2861510015&rft_id=info:pmid/&rfr_iscdi=true