Sensitivity as a Complexity Measure for Sequence Classification Tasks

We introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of Boolean function sensitivity. The sensitivity of a function, given a distribution over input sequences, quantifies the number of disjoint su...

Full description

Saved in:
Bibliographic Details
Published in:Transactions of the Association for Computational Linguistics Vol. 9; pp. 891 - 908
Main Authors: Hahn, Michael, Jurafsky, Dan, Futrell, Richard
Format: Journal Article
Language:English
Published: One Rogers Street, Cambridge, MA 02142-1209, USA MIT Press 18.08.2021
MIT Press Journals, The
The MIT Press
Subjects:
ISSN:2307-387X, 2307-387X
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract We introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of Boolean function sensitivity. The sensitivity of a function, given a distribution over input sequences, quantifies the number of disjoint subsets of the input sequence that can each be individually changed to change the output. We argue that standard sequence classification methods are biased towards learning low-sensitivity functions, so that tasks requiring high sensitivity are more difficult. To that end, we show analytically that simple lexical classifiers can only express functions of bounded sensitivity, and we show empirically that low-sensitivity functions are easier to learn for LSTMs. We then estimate sensitivity on 15 NLP tasks, finding that sensitivity is higher on challenging tasks collected in GLUE than on simple text classification tasks, and that sensitivity predicts the performance both of simple lexical classifiers and of vanilla BiLSTMs without pretrained contextualized embeddings. Within a task, sensitivity predicts which inputs are hard for such simple models. Our results suggest that the success of massively pretrained contextual representations stems in part because they provide representations from which information can be extracted by low-sensitivity decoders.
AbstractList We introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of Boolean function sensitivity. The sensitivity of a function, given a distribution over input sequences, quantifies the number of disjoint subsets of the input sequence that can each be individually changed to change the output. We argue that standard sequence classification methods are biased towards learning low-sensitivity functions, so that tasks requiring high sensitivity are more difficult. To that end, we show analytically that simple lexical classifiers can only express functions of bounded sensitivity, and we show empirically that low-sensitivity functions are easier to learn for LSTMs. We then estimate sensitivity on 15 NLP tasks, finding that sensitivity is higher on challenging tasks collected in GLUE than on simple text classification tasks, and that sensitivity predicts the performance both of simple lexical classifiers and of vanilla BiLSTMs without pretrained contextualized embeddings. Within a task, sensitivity predicts which inputs are hard for such simple models. Our results suggest that the success of massively pretrained contextual representations stems in part because they provide representations from which information can be extracted by low-sensitivity decoders.
AbstractWe introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of Boolean function sensitivity. The sensitivity of a function, given a distribution over input sequences, quantifies the number of disjoint subsets of the input sequence that can each be individually changed to change the output. We argue that standard sequence classification methods are biased towards learning low-sensitivity functions, so that tasks requiring high sensitivity are more difficult. To that end, we show analytically that simple lexical classifiers can only express functions of bounded sensitivity, and we show empirically that low-sensitivity functions are easier to learn for LSTMs. We then estimate sensitivity on 15 NLP tasks, finding that sensitivity is higher on challenging tasks collected in GLUE than on simple text classification tasks, and that sensitivity predicts the performance both of simple lexical classifiers and of vanilla BiLSTMs without pretrained contextualized embeddings. Within a task, sensitivity predicts which inputs are hard for such simple models. Our results suggest that the success of massively pretrained contextual representations stems in part because they provide representations from which information can be extracted by low-sensitivity decoders.
Author Jurafsky, Dan
Hahn, Michael
Futrell, Richard
Author_xml – sequence: 1
  givenname: Michael
  surname: Hahn
  fullname: Hahn, Michael
  email: mhahn2@stanford.edu
  organization: Stanford University, United States. mhahn2@stanford.edu
– sequence: 2
  givenname: Dan
  surname: Jurafsky
  fullname: Jurafsky, Dan
  email: jurafsky@stanford.edu
  organization: Stanford University, United States. jurafsky@stanford.edu
– sequence: 3
  givenname: Richard
  surname: Futrell
  fullname: Futrell, Richard
  email: rfutrell@uci.edu
  organization: University of California, Irvine, United States. rfutrell@uci.edu
BookMark eNp1kE9vEzEQxS1UJErpjQ-wEhcOpIz_rNd74ICiQisVcWiRuFlj7xg5bNbBdqqWT8-mASmqymns0e-9eTMv2dGUJmLsNYczzrV4X9GPFi2AAvmMHQsJ3UKa7vvRwfsFOy1lBQDccANaHLPza5pKrPE21vsGS4PNMq03I93t_l8IyzZTE1JurunXliZPzXLEUmKIHmtMU3OD5Wd5xZ4HHAud_q0n7Nun85vlxeLq6-fL5cerhVetqgvtqNMuCBOUo-C4EdpxAcgV10pqBx76Xre9AyNpaMGTdr6XBg2C9hzkCbvc-w4JV3aT4xrzvU0Y7UMj5R8Wc41-JNu3hoJUqh3coHw3OCQUQ-elgSF0Qsxeb_Zem5zm1Uq1q7TN0xzfCjMPNart9EyJPeVzKiVTsD7Wh81rxjhaDnZ3fXt4_Vn07pHoX9T_4B_2-DoehNght72VSmveWwFCWphrb3_HzWP92yf0T476A-1aqn0
CitedBy_id crossref_primary_10_1007_s42421_024_00109_x
crossref_primary_10_1016_j_jbi_2025_104881
Cites_doi 10.1162/neco.1997.9.8.1735
10.3115/v1/D14-1181
10.1145/1014052.1014073
10.18653/v1/2020.acl-main.24
10.1017/S1351324909990209
10.1162/tacl_a_00115
10.18653/v1/D17-1215
10.1017/CBO9781139814782
10.18653/v1/D18-1151
10.18653/v1/S18-2023
10.1109/TIT.1956.1056813
10.1109/SFCS.1988.21923
10.18653/v1/N19-1004
10.1016/0020-0190(96)00105-6
10.18653/v1/D16-1264
10.3115/1073336.1073357
10.18653/v1/2020.acl-main.768
10.3115/1219840.1219855
10.1016/j.neucom.2006.01.025
10.1137/0220062
10.18653/v1/W19-3901
10.18653/v1/D18-1546
10.1162/tacl_a_00306
10.18653/v1/P19-1449
10.18653/v1/N18-1101
10.18653/v1/N18-2017
10.18653/v1/P18-1031
10.3115/v1/D14-1162
10.1007/978-1-4757-3860-5
10.1007/s10579-005-7880-9
10.1007/978-3-030-36708-4_22
10.18653/v1/2020.emnlp-demos.6
10.3115/1218955.1218990
10.1162/tacl_a_00290
10.18653/v1/N18-1202
10.1016/S0010-0277(98)00034-1
10.18653/v1/2020.acl-demos.14
10.18653/v1/2020.findings-emnlp.117
10.18653/v1/W18-3012
10.18653/v1/2020.acl-demos.10
10.18653/v1/S17-2001
ContentType Journal Article
Copyright 2021. This work is published under https://creativecommons.org/licenses/by/4.0/legalcode (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Copyright_xml – notice: 2021. This work is published under https://creativecommons.org/licenses/by/4.0/legalcode (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
DBID AAYXX
CITATION
7T9
8FE
8FG
ABUWG
AFKRA
ALSLI
ARAPS
AZQEC
BENPR
BGLVJ
CCPQU
CPGLG
CRLPW
DWQXO
GNUQQ
HCIFZ
JQ2
K7-
P5Z
P62
PHGZM
PHGZT
PIMPY
PKEHL
PQEST
PQGLB
PQQKQ
PQUKI
PRINS
PRQQA
DOA
DOI 10.1162/tacl_a_00403
DatabaseName CrossRef
Linguistics and Language Behavior Abstracts (LLBA)
ProQuest SciTech Collection
ProQuest Technology Collection
ProQuest Central (Alumni)
ProQuest Central UK/Ireland
Social Science Premium Collection
Advanced Technologies & Computer Science Collection
ProQuest Central Essentials
ProQuest Central
Technology Collection
ProQuest One Community College
Linguistics collection.
Linguistics Database
ProQuest Central
ProQuest Central Student
SciTech Premium Collection
ProQuest Computer Science Collection
Computer Science Database
Advanced Technologies & Aerospace Database
ProQuest Advanced Technologies & Aerospace Collection
ProQuest Central Premium
ProQuest One Academic
Publicly Available Content Database
ProQuest One Academic Middle East (New)
ProQuest One Academic Eastern Edition (DO NOT USE)
One Applied & Life Sciences
ProQuest One Academic (retired)
ProQuest One Academic UKI Edition
ProQuest Central China
ProQuest One Social Sciences
DOAJ Directory of Open Access Journals
DatabaseTitle CrossRef
Publicly Available Content Database
Computer Science Database
ProQuest Central Student
Technology Collection
ProQuest One Academic Middle East (New)
ProQuest Advanced Technologies & Aerospace Collection
ProQuest Central Essentials
ProQuest Computer Science Collection
ProQuest Central (Alumni Edition)
SciTech Premium Collection
ProQuest One Community College
ProQuest Central China
ProQuest Central
ProQuest One Applied & Life Sciences
Linguistics Collection
ProQuest Central Korea
ProQuest Central (New)
Advanced Technologies & Aerospace Collection
Social Science Premium Collection
ProQuest One Social Sciences
ProQuest One Academic Eastern Edition
Linguistics and Language Behavior Abstracts (LLBA)
ProQuest Technology Collection
ProQuest SciTech Collection
Advanced Technologies & Aerospace Database
ProQuest One Academic UKI Edition
Linguistics Database
ProQuest One Academic
ProQuest One Academic (New)
DatabaseTitleList CrossRef
Publicly Available Content Database


Database_xml – sequence: 1
  dbid: DOA
  name: DOAJ Directory of Open Access Journals
  url: https://www.doaj.org/
  sourceTypes: Open Website
– sequence: 2
  dbid: PIMPY
  name: Publicly Available Content Database
  url: http://search.proquest.com/publiccontent
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
EISSN 2307-387X
EndPage 908
ExternalDocumentID oai_doaj_org_article_958ef3445dbd4c7dbaea2d7c380df722
10_1162_tacl_a_00403
tacl_a_00403.pdf
GroupedDBID AAFWJ
AFPKN
ALMA_UNASSIGNED_HOLDINGS
EBS
GROUPED_DOAJ
JMNJE
M~E
OJV
OK1
RMI
AAYXX
ABUWG
AFFHD
AFKRA
ALSLI
ARAPS
BENPR
BGLVJ
CCPQU
CITATION
CPGLG
CRLPW
DWQXO
HCIFZ
K7-
PHGZM
PHGZT
PIMPY
PQGLB
PRQQA
7T9
8FE
8FG
AZQEC
GNUQQ
JQ2
P62
PKEHL
PQEST
PQQKQ
PQUKI
PRINS
ID FETCH-LOGICAL-c454t-6be76bf28f4befb1826b120a1416436b0c099659b083ed50ce6bc938a8a06c103
IEDL.DBID BENPR
ISICitedReferencesCount 10
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000751952200053&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 2307-387X
IngestDate Fri Oct 03 12:52:10 EDT 2025
Sat Nov 08 19:51:57 EST 2025
Sat Nov 29 05:34:28 EST 2025
Tue Nov 18 22:40:09 EST 2025
Sat Sep 30 12:10:40 EDT 2023
Sun Oct 01 15:05:32 EDT 2023
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c454t-6be76bf28f4befb1826b120a1416436b0c099659b083ed50ce6bc938a8a06c103
Notes 2021
ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
OpenAccessLink https://www.proquest.com/docview/2893884576?pq-origsite=%requestingapplication%
PQID 2893884576
PQPubID 6535866
PageCount 18
ParticipantIDs proquest_journals_2893884576
crossref_primary_10_1162_tacl_a_00403
doaj_primary_oai_doaj_org_article_958ef3445dbd4c7dbaea2d7c380df722
crossref_citationtrail_10_1162_tacl_a_00403
mit_journals_taclv9_346619_2023_09_29_zip_tacl_a_00403
mit_journals_10_1162_tacl_a_00403
PublicationCentury 2000
PublicationDate 2021-08-18
PublicationDateYYYYMMDD 2021-08-18
PublicationDate_xml – month: 08
  year: 2021
  text: 2021-08-18
  day: 18
PublicationDecade 2020
PublicationPlace One Rogers Street, Cambridge, MA 02142-1209, USA
PublicationPlace_xml – name: One Rogers Street, Cambridge, MA 02142-1209, USA
– name: Cambridge
PublicationTitle Transactions of the Association for Computational Linguistics
PublicationYear 2021
Publisher MIT Press
MIT Press Journals, The
The MIT Press
Publisher_xml – name: MIT Press
– name: MIT Press Journals, The
– name: The MIT Press
References O’Donnell (2021081914590098500_bib43) 2014
Howard (2021081914590098500_bib23) 2018
Cao (2021081914590098500_bib3) 2019
Poliak (2021081914590098500_bib51) 2018
Jeretic (2021081914590098500_bib26) 2020
Horne (2021081914590098500_bib22) 1994
Gauthier (2021081914590098500_bib15) 2020
Rajpurkar (2021081914590098500_bib56) 2016
Wang (2021081914590098500_bib63) 2019
Gururangan (2021081914590098500_bib17) 2018
Pennington (2021081914590098500_bib48) 2014
Radford (2021081914590098500_bib54) 2019; 1
Qi (2021081914590098500_bib52) 2018
Merrill (2021081914590098500_bib37) 2019
Qi (2021081914590098500_bib53) 2020
Peters (2021081914590098500_bib49) 2018
Kaushik (2021081914590098500_bib29) 2020
Nivre (2021081914590098500_bib41) 2016
Hatami (2021081914590098500_bib20) 2010; 4
Chomsky (2021081914590098500_bib6) 1956; 2
Szegedy (2021081914590098500_bib59) 2014
Dolan (2021081914590098500_bib10) 2005
Pang (2021081914590098500_bib47) 2005
Minsky (2021081914590098500_bib38) 1969
Levesque (2021081914590098500_bib32) 2012
Wiebe (2021081914590098500_bib65) 2005; 39
Hochreiter (2021081914590098500_bib21) 1997; 9
Hu (2021081914590098500_bib25) 2004
Datta (2021081914590098500_bib8) 2020
Hale (2021081914590098500_bib19) 2001
Ott (2021081914590098500_bib44) 2019
Valle-Perez (2021081914590098500_bib60) 2019
Hahn (2021081914590098500_bib18) 2020; 8
Novak (2021081914590098500_bib42) 2018
Franco (2021081914590098500_bib12) 2006; 70
Kim (2021081914590098500_bib31) 2014
Gibson (2021081914590098500_bib16) 1998; 68
Li (2021081914590098500_bib33) 1993
Silveira (2021081914590098500_bib57) 2014
Wolf (2021081914590098500_bib68) 2019
Devlin (2021081914590098500_bib9) 2019
Dagan (2021081914590098500_bib7) 2009; 15
Gardner (2021081914590098500_bib14) 2020
Wang (2021081914590098500_bib62) 2019
Jia (2021081914590098500_bib27) 2017
Warstadt (2021081914590098500_bib64) 2019; 7
Kaushik (2021081914590098500_bib30) 2018
Petrov (2021081914590098500_bib50) 2012
Wieting (2021081914590098500_bib66) 2016
Arora (2021081914590098500_bib1) 2017
Futrell (2021081914590098500_bib13) 2019
Marvin (2021081914590098500_bib36) 2018
Liao (2021081914590098500_bib34) 2020
Xu (2021081914590098500_bib69) 2019
Nisan (2021081914590098500_bib40) 1991; 20
Chelba (2021081914590098500_bib5) 2014
Vaswani (2021081914590098500_bib61) 2017
Williams (2021081914590098500_bib67) 2018
Nangia (2021081914590098500_bib39) 2019
Hu (2021081914590098500_bib24) 2020; 3
Palma (2021081914590098500_bib45) 2019
Yang (2021081914590098500_bib70) 2019
Bernasconi (2021081914590098500_bib2) 1996; 59
Cer (2021081914590098500_bib4) 2017
Rahaman (2021081914590098500_bib55) 2019
Kahn (2021081914590098500_bib28) 1988
Pang (2021081914590098500_bib46) 2004
Socher (2021081914590098500_bib58) 2013
Ethayarajh (2021081914590098500_bib11) 2018
Linzen (2021081914590098500_bib35) 2016; 4
References_xml – volume: 9
  start-page: 1735
  issue: 8
  year: 1997
  ident: 2021081914590098500_bib21
  article-title: Long short-term memory
  publication-title: Neural Computation
  doi: 10.1162/neco.1997.9.8.1735
– start-page: 1746
  volume-title: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)
  year: 2014
  ident: 2021081914590098500_bib31
  article-title: Convolutional neural networks for sentence classification
  doi: 10.3115/v1/D14-1181
– year: 2019
  ident: 2021081914590098500_bib3
  article-title: Towards understanding the spectral bias of deep learning
  publication-title: arXiv preprint arXiv:1912.01198
– start-page: 168
  volume-title: Proceedings of the Tenth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
  year: 2004
  ident: 2021081914590098500_bib25
  article-title: Mining and summarizing customer reviews
  doi: 10.1145/1014052.1014073
– start-page: 263
  volume-title: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
  year: 2020
  ident: 2021081914590098500_bib34
  article-title: Probabilistically masked language model capable of autoregressive generation in arbitrary word order
  doi: 10.18653/v1/2020.acl-main.24
– volume: 15
  issue: 4
  year: 2009
  ident: 2021081914590098500_bib7
  article-title: Recognizing textual entailment: Rational, evaluation and approaches
  publication-title: Natural Language Engineering
  doi: 10.1017/S1351324909990209
– volume: 4
  start-page: 521
  year: 2016
  ident: 2021081914590098500_bib35
  article-title: Assessing the ability of LSTMs to learn syntax-sensitive dependencies
  publication-title: Transactions of the Association for Computational Linguistics
  doi: 10.1162/tacl_a_00115
– start-page: 2021
  volume-title: Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing
  year: 2017
  ident: 2021081914590098500_bib27
  article-title: Adversarial examples for evaluating reading comprehension systems
  doi: 10.18653/v1/D17-1215
– volume-title: Analysis of Boolean Functions
  year: 2014
  ident: 2021081914590098500_bib43
  doi: 10.1017/CBO9781139814782
– start-page: 552
  volume-title: KR’12 Proceedings of the Thirteenth International Conference on Principles of Knowledge Representation and Reasoning
  year: 2012
  ident: 2021081914590098500_bib32
  article-title: The winograd schema challenge
– year: 2018
  ident: 2021081914590098500_bib36
  article-title: Targeted syntactic evaluation of language models
  publication-title: arXiv preprint arXiv:1808.09031
  doi: 10.18653/v1/D18-1151
– volume-title: ICLR 2020: Eighth International Conference on Learning Representations
  year: 2020
  ident: 2021081914590098500_bib29
  article-title: Learning the difference that makes a difference with counterfactually-augmented data
– start-page: 180
  volume-title: Proceedings of the Seventh Joint Conference on Lexical and Computational Semantics
  year: 2018
  ident: 2021081914590098500_bib51
  article-title: Hypothesis only baselines in natural language inference
  doi: 10.18653/v1/S18-2023
– volume: 4
  start-page: 1
  year: 2010
  ident: 2021081914590098500_bib20
  article-title: Variations on the sensitivity conjecture
  publication-title: Theory of Computing
– start-page: 359
  volume-title: Advances in Neural Information Processing Systems
  year: 1994
  ident: 2021081914590098500_bib22
  article-title: Bounds on the complexity of recurrent neural network implementations of finite state machines
– volume: 2
  start-page: 113
  issue: 3
  year: 1956
  ident: 2021081914590098500_bib6
  article-title: Three models for the description of language
  publication-title: IEEE Transactions on Information Theory
  doi: 10.1109/TIT.1956.1056813
– start-page: 160
  volume-title: Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies
  year: 2018
  ident: 2021081914590098500_bib52
  article-title: Universal dependency parsing from scratch
– volume-title: 6th International Conference on Learning Representations, ICLR 2018, Vancouver, BC, Canada, April 30 – May 3, 2018, Conference Track Proceedings
  year: 2018
  ident: 2021081914590098500_bib42
  article-title: Sensitivity and generalization in neural networks: an empirical study
– volume-title: Proceedings of the Third International Workshop on Paraphrasing, IWP@IJCNLP 2005, Jeju Island, Korea, October 2005, 2005
  year: 2005
  ident: 2021081914590098500_bib10
  article-title: Automatically constructing a corpus of sentential paraphrases
– volume-title: Tenth International Conference on Language Resources and Evaluation (LREC 2016)
  year: 2016
  ident: 2021081914590098500_bib41
  article-title: Universal dependencies v1: A multilingual treebank collection
– volume-title: 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14–16, 2014, Conference Track Proceedings
  year: 2014
  ident: 2021081914590098500_bib59
  article-title: Intriguing properties of neural networks
– start-page: 2635
  volume-title: INTERSPEECH
  year: 2014
  ident: 2021081914590098500_bib5
  article-title: One billion word benchmark for measuring progress in statistical language modeling
– start-page: 1962
  volume-title: Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8–14, 2019, Vancouver, BC, Canada
  year: 2019
  ident: 2021081914590098500_bib45
  article-title: Random deep neural networks are biased towards simple functions
– start-page: 68
  volume-title: [Proceedings 1988] 29th Annual Symposium on Foundations of Computer Science
  year: 1988
  ident: 2021081914590098500_bib28
  article-title: The influence of variables on Boolean functions
  doi: 10.1109/SFCS.1988.21923
– start-page: 32
  volume-title: 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)
  year: 2019
  ident: 2021081914590098500_bib13
  article-title: Neural language models as psycholinguistic subjects: Representations of syntactic state
  doi: 10.18653/v1/N19-1004
– volume-title: 4th International Conference on Learning Representations, ICLR 2016, San Juan, Puerto Rico, May 2–4, 2016, Conference Track Proceedings
  year: 2016
  ident: 2021081914590098500_bib66
  article-title: Towards universal paraphrastic sentence embeddings
– volume: 59
  start-page: 151
  issue: 3
  year: 1996
  ident: 2021081914590098500_bib2
  article-title: Sensitivity vs. block sensitivity (an average-case study)
  publication-title: Information Processing Letters
  doi: 10.1016/0020-0190(96)00105-6
– start-page: 2383
  volume-title: Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing
  year: 2016
  ident: 2021081914590098500_bib56
  article-title: Squad: 100,000+ questions for machine comprehension of text
  doi: 10.18653/v1/D16-1264
– start-page: 1
  volume-title: Proceedings of the Second Meeting of the North American Chapter of the Association for Computational Linguistics and Language Technologies
  year: 2001
  ident: 2021081914590098500_bib19
  article-title: A probabilistic Earley parser as a psycholinguistic model
  doi: 10.3115/1073336.1073357
– start-page: 8690
  volume-title: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL 2020, Online, July 5–10, 2020
  year: 2020
  ident: 2021081914590098500_bib26
  article-title: Are natural language inference models IMPPRESsive? Learning IMPlicature and PRESupposition
  doi: 10.18653/v1/2020.acl-main.768
– start-page: 115
  volume-title: Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL’05)
  year: 2005
  ident: 2021081914590098500_bib47
  article-title: Seeing stars: Exploiting class relationships for sentiment categorization with respect to rating scales
  doi: 10.3115/1219840.1219855
– start-page: 2897
  volume-title: Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14)
  year: 2014
  ident: 2021081914590098500_bib57
  article-title: A gold standard dependency corpus for English
– volume-title: ICLR 2017: International Conference on Learning Representations 2017
  year: 2017
  ident: 2021081914590098500_bib1
  article-title: A simple but tough-to-beat baseline for sentence embeddings
– volume: 70
  start-page: 351
  issue: 1
  year: 2006
  ident: 2021081914590098500_bib12
  article-title: Generalization ability of boolean functions implemented in feedforward neural networks
  publication-title: Neurocomputing
  doi: 10.1016/j.neucom.2006.01.025
– volume: 1
  start-page: 9
  issue: 8
  year: 2019
  ident: 2021081914590098500_bib54
  article-title: Language models are unsupervised multitask learners
  publication-title: OpenAI Blog
– start-page: 5998
  volume-title: Advances in Neural Information Processing Systems
  year: 2017
  ident: 2021081914590098500_bib61
  article-title: Attention is all you need
– volume: 20
  start-page: 999
  issue: 6
  year: 1991
  ident: 2021081914590098500_bib40
  article-title: CREW PRAMs and decision trees
  publication-title: SIAM Journal on Computing
  doi: 10.1137/0220062
– year: 2019
  ident: 2021081914590098500_bib37
  article-title: Sequential neural networks as automata
  publication-title: arXiv preprint arXiv:1906.01615
  doi: 10.18653/v1/W19-3901
– start-page: 5010
  volume-title: EMNLP 2018: 2018 Conference on Empirical Methods in Natural Language Processing
  year: 2018
  ident: 2021081914590098500_bib30
  article-title: How much reading does reading comprehension require? A critical investigation of popular benchmarks
  doi: 10.18653/v1/D18-1546
– volume-title: Perceptrons: An Introduction to Computational Geometry
  year: 1969
  ident: 2021081914590098500_bib38
– start-page: 48
  volume-title: NAACL-HLT 2019: Annual Conference of the North American Chapter of the Association for Computational Linguistics
  year: 2019
  ident: 2021081914590098500_bib44
  article-title: fairseq: A fast, extensible toolkit for sequence modeling
– start-page: 4171
  volume-title: NAACL-HLT 2019: Annual Conference of the North American Chapter of the Association for Computational Linguistics
  year: 2019
  ident: 2021081914590098500_bib9
  article-title: BERT: Pre-training of deep bidirectional transformers for language understanding
– volume: 8
  start-page: 156
  year: 2020
  ident: 2021081914590098500_bib18
  article-title: Theoretical limitations of self-attention in neural sequence models
  publication-title: Transactions of the Association for Computational Linguistics
  doi: 10.1162/tacl_a_00306
– start-page: 4566
  volume-title: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics
  year: 2019
  ident: 2021081914590098500_bib39
  article-title: Human vs. muppet: A conservative estimate of human performance on the glue benchmark
  doi: 10.18653/v1/P19-1449
– start-page: 1112
  volume-title: NAACL HLT 2018: 16th Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
  year: 2018
  ident: 2021081914590098500_bib67
  article-title: A broad-coverage challenge corpus for sentence understanding through inference
  doi: 10.18653/v1/N18-1101
– volume: 3
  start-page: 382
  issue: 1
  year: 2020
  ident: 2021081914590098500_bib24
  article-title: A closer look at the performance of neural language models on reflexive anaphor licensing
  publication-title: Proceedings of the Society for Computation in Linguistics
– start-page: 107
  volume-title: NAACL HLT 2018: 16th Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
  year: 2018
  ident: 2021081914590098500_bib17
  article-title: Annotation artifacts in natural language inference data
  doi: 10.18653/v1/N18-2017
– start-page: 328
  volume-title: ACL 2018: 56th Annual Meeting of the Association for Computational Linguistics
  year: 2018
  ident: 2021081914590098500_bib23
  article-title: Universal language model fine-tuning for text classification
  doi: 10.18653/v1/P18-1031
– start-page: 2089
  volume-title: Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC-2012)
  year: 2012
  ident: 2021081914590098500_bib50
  article-title: A universal part-of-speech tagset
– start-page: 1532
  volume-title: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)
  year: 2014
  ident: 2021081914590098500_bib48
  article-title: GloVe: Global vectors for word representation
  doi: 10.3115/v1/D14-1162
– volume-title: An Introduction to Kolmogorov Complexity and its Applications
  year: 1993
  ident: 2021081914590098500_bib33
  doi: 10.1007/978-1-4757-3860-5
– volume: 39
  start-page: 165
  issue: 2
  year: 2005
  ident: 2021081914590098500_bib65
  article-title: Annotating expressions of opinions and emotions in language
  publication-title: Language Resources and Evaluation
  doi: 10.1007/s10579-005-7880-9
– start-page: 1631
  volume-title: Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing
  year: 2013
  ident: 2021081914590098500_bib58
  article-title: Recursive deep models for semantic compositionality over a sentiment treebank
– start-page: 264
  volume-title: International Conference on Neural Information Processing
  year: 2019
  ident: 2021081914590098500_bib69
  article-title: Training behavior of deep neural network in frequency domain
  doi: 10.1007/978-3-030-36708-4_22
– year: 2019
  ident: 2021081914590098500_bib68
  article-title: Huggingface’s transformers: State-of-the-art natural language processing
  publication-title: arXiv preprint arXiv:1910.03771
  doi: 10.18653/v1/2020.emnlp-demos.6
– start-page: 5301
  volume-title: ICML 2019: Thirty-sixth International Conference on Machine Learning
  year: 2019
  ident: 2021081914590098500_bib55
  article-title: On the spectral bias of neural networks
– start-page: 271
  volume-title: Proceedings of the 42nd Meeting of the Association for Computational Linguistics (ACL’04), Main Volume
  year: 2004
  ident: 2021081914590098500_bib46
  article-title: A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts
  doi: 10.3115/1218955.1218990
– volume: 7
  start-page: 625
  year: 2019
  ident: 2021081914590098500_bib64
  article-title: Neural network acceptability judgments
  publication-title: Transactions of the Association for Computational Linguistics
  doi: 10.1162/tacl_a_00290
– year: 2020
  ident: 2021081914590098500_bib8
  article-title: Geometry matters: Exploring language examples at the decision boundary
  publication-title: CoRR
– volume-title: ICLR 2019: 7th International Conference on Learning Representations
  year: 2019
  ident: 2021081914590098500_bib60
  article-title: Deep learning generalizes because the parameter-function map is biased towards simple functions
– start-page: 2227
  volume-title: NAACL HLT 2018: 16th Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
  year: 2018
  ident: 2021081914590098500_bib49
  article-title: Deep contextualized word representations
  doi: 10.18653/v1/N18-1202
– volume: 68
  start-page: 1
  issue: 1
  year: 1998
  ident: 2021081914590098500_bib16
  article-title: Linguistic complexity: Locality of syntactic dependencies
  publication-title: Cognition
  doi: 10.1016/S0010-0277(98)00034-1
– start-page: 101
  volume-title: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: System Demonstrations, ACL 2020, Online, July 5–10, 2020
  year: 2020
  ident: 2021081914590098500_bib53
  article-title: Stanza: A Python natural language processing toolkit for many human languages
  doi: 10.18653/v1/2020.acl-demos.14
– start-page: 1307
  volume-title: Findings of the Association for Computational Linguistics: EMNLP 2020
  year: 2020
  ident: 2021081914590098500_bib14
  article-title: Evaluating NLP models via contrast sets
  doi: 10.18653/v1/2020.findings-emnlp.117
– start-page: 3266
  volume-title: Advances in Neural Information Processing Systems
  year: 2019
  ident: 2021081914590098500_bib62
  article-title: SuperGLUE: A stickier benchmark for general-purpose language understanding systems
– start-page: 5753
  volume-title: NeurIPS 2019: Thirty-third Conference on Neural Information Processing Systems
  year: 2019
  ident: 2021081914590098500_bib70
  article-title: XLNet: Generalized autoregressive pretraining for language understanding
– start-page: 91
  volume-title: Proceedings of The Third Workshop on Representation Learning for NLP
  year: 2018
  ident: 2021081914590098500_bib11
  article-title: Unsupervised random walk sentence embeddings: A strong but simple baseline
  doi: 10.18653/v1/W18-3012
– volume-title: Proceedings of the Association for Computational Linguistics: System Demonstrations (ACL 2020)
  year: 2020
  ident: 2021081914590098500_bib15
  article-title: SyntaxGym: An online platform for targeted evaluation of language models
  doi: 10.18653/v1/2020.acl-demos.10
– volume-title: ICLR 2019: 7th International Conference on Learning Representations
  year: 2019
  ident: 2021081914590098500_bib63
  article-title: GLUE: A multi-task benchmark and analysis platform for natural language understanding
  doi: 10.1162/tacl_a_00290
– start-page: 1
  volume-title: Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017)
  year: 2017
  ident: 2021081914590098500_bib4
  article-title: Semeval-2017 task 1: Semantic textual similarity multilingual and crosslingual focused evaluation
  doi: 10.18653/v1/S17-2001
SSID ssj0001818062
Score 2.348562
Snippet We introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of...
AbstractWe introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the...
SourceID doaj
proquest
crossref
mit
SourceType Open Website
Aggregation Database
Enrichment Source
Index Database
Publisher
StartPage 891
SubjectTerms Boolean functions
Classification
Classifiers
Decoders
Natural language processing
Representations
Sequences
Subsets
Task complexity
Task performance
SummonAdditionalLinks – databaseName: DOAJ Directory of Open Access Journals
  dbid: DOA
  link: http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV3NS8MwFA8yPHgRRcXplAh6kmKapGlyVNnwoEPwA28hnzCcU9a6g3-9SdrpVMSLp9L2pWnfS_LeC6-_HwCHlntFw9qbae5wRp0WmdAmzxjyjGrDCEl0QPeX5XDIHx7E9QLVV6wJa-CBG8WdiII7TygtrLbUlFYrp7AtDeHI-hKn1ReVYiGZSrsr8RdmhueV7gyf1MqMpZJx0JIvPihB9QfP8jSqf6zHyckM1sBqGx3C0-at1sGSm2yA_k0sMm9YHqCqoIJxEkcgy3B-1ezxwRB7wpu2LBompstYA5TUDm9V9VhtgrtB__b8ImvZDzJDC1pnTLuSaY-5p9p5HfMAnWOk8hBCUcI0MiG4Y4XQIYhytkDGMW0E4YorxEyOyBboTJ4nbhtAZYPmsSMkV5iWNlc-xwozbsKTlLeoC47n-pCmhQaPDBVjmVIEhuWi9rrg6EP6pYHE-EXuLKr2QyYCWacLwbyyNa_8y7xdcBAMI9uJVf3SEfsiE-_NhCQ0xB5CRn54icJRyLfRy7eGvbm9P1uHDJRwTkMetvMfH7ALVnAsg4kourwHOvX01e2BZTOrR9V0P43ad9oW81s
  priority: 102
  providerName: Directory of Open Access Journals
Title Sensitivity as a Complexity Measure for Sequence Classification Tasks
URI https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00403
https://www.proquest.com/docview/2893884576
https://doaj.org/article/958ef3445dbd4c7dbaea2d7c380df722
Volume 9
WOSCitedRecordID wos000751952200053&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVAON
  databaseName: DOAJ Directory of Open Access Journals
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 20241231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: DOA
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://www.doaj.org/
  providerName: Directory of Open Access Journals
– providerCode: PRVHPJ
  databaseName: ROAD: Directory of Open Access Scholarly Resources
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: M~E
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://road.issn.org
  providerName: ISSN International Centre
– providerCode: PRVPQU
  databaseName: Advanced Technologies & Aerospace Database
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: P5Z
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/hightechjournals
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Computer Science Database
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: K7-
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: http://search.proquest.com/compscijour
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Linguistics Database
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: CRLPW
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://search.proquest.com/linguistics
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: ProQuest Central
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: BENPR
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: https://www.proquest.com/central
  providerName: ProQuest
– providerCode: PRVPQU
  databaseName: Publicly Available Content Database
  customDbUrl:
  eissn: 2307-387X
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0001818062
  issn: 2307-387X
  databaseCode: PIMPY
  dateStart: 20130101
  isFulltext: true
  titleUrlDefault: http://search.proquest.com/publiccontent
  providerName: ProQuest
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1Lb9QwELag5cCFggCxUKogwQlFdWzHcU5Vi7YCQVcRLWjhYvmVakXZXTZpD_31nfF6uxRULlwSJR7nYY_H48nk-wh57VVrBNje3KrAchFsndfWFbmkrRTWSc4jHdDXT9VopMbjukkBty6lVa5sYjTUfuYwRr4LCwOulAD3eG_-K0fWKPy6mig07pJNRCoDPd88GI6az-soC_7KHFlFMeEZgWTHq-x3yXZ748600ajI_Ma8FOH7Ybb5Oen_stFx4jnc-t9HfkgeJJcz21_qyCNyJ0wfk-ExZq4vqSMy02UmQ8uA6JhwfLQMHGbg0GbHKdc6i_SZmFgU-zI7Md2P7gn5cjg8efc-T5QKuROl6HNpQyVty1QrbGgtLi5swagpwC8TXFrqwGOUZW3BMwu-pC5I6-AtjDJUuoLyp2RjOpuGZyQzHrqTBc4Lw0TlC9MWzDCpHFzJtJ4OyNtVg2qX8MaR9uJMx3WHZPr35h-QN9fS8yXOxi1yB9g31zKIjh1PzBanOg02XZcqtFyI0lsvXOWtCYb5ynFFfVsxNiCvoGd1Gq3dLTeSN2Sw7KLWXIBDU2skndcU9rW-nMz_qLi90oZ17bUqPP938Qtyn2HWDILuqm2y0S_Ow0tyz130k26xk9R7J0YOYPuxymHblN-hpPlw1Hy7AiJKCDg
linkProvider ProQuest
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMw1V1Lb9QwELZKiwQXHgLEQgEjUS4oqmM7jnPgAKVVq25XFV1gb65fQSvK7rIJRfCj-I3MZJMuBZVbD5yixGNbicczn53xfIQ8C7q0Emxv4nTkiYyuSArn00SxUknnlRANHdD7fj4Y6NGoOFwhP7uzMBhW2dnExlCHqcc98k1YGAitJcDjNoJyP37_Buuz6uXeGxjMDc53todbu0lLIZB4mck6US7mypVcl9LF0iGYdilnNgUcIoVyzANCUlnhAInEkDEflfPQkdWWKZ8yAe0-n31JkKUK_-a2lB1XyJpWRQ6GYG3rbf_ww3JXB49ONyymGGCNiWtHXbS94pu19SfGGpw44pwfbOgCwLt9Htd_-YTG0e3c_N8-0S1yo4XU9NViDtwmK3Fyh2wfYWT-ghqD2opaipYPs3_C_cFiY5QCYKdHbSw5behBMXCq0VU6tNWn6i55dykvc4-sTqaTeJ9QG0BdeRQitVzmIbVlyi1X2kNLtgysR150A2h8m08daT1OTLOuUtz8Ptw9snEmPVvkEblA7jXqwpkMZv9uHkznH01rTEyR6VgKKbPggvR5cDZaHnIvNAtlznmPPAVNMq01qi7oSJ2TwbLTwggJgK0wHKCcYXAtzI_x7I-K6532LWsvVe_Bv4ufkGu7w4O-6e8N9h-S6xwjhDDBsF4nq_X8a3xErvrTelzNH7dTi5Ljy9beXwjDX9E
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Sensitivity+as+a+Complexity+Measure+for+Sequence+Classification+Tasks&rft.jtitle=Transactions+of+the+Association+for+Computational+Linguistics&rft.au=Hahn%2C+Michael&rft.au=Jurafsky%2C+Dan&rft.au=Futrell%2C+Richard&rft.date=2021-08-18&rft.pub=MIT+Press+Journals%2C+The&rft.issn=2307-387X&rft.eissn=2307-387X&rft.volume=9&rft.spage=891&rft_id=info:doi/10.1162%2Ftacl_a_00403
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2307-387X&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2307-387X&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2307-387X&client=summon