@inproceedings{lin-etal-2019-open,
title = "Open Sesame: Getting inside {BERT}{'}s Linguistic Knowledge",
author = "Lin, Yongjie and
Tan, Yi Chern and
Frank, Robert",
editor = "Linzen, Tal and
Chrupa{\l}a, Grzegorz and
Belinkov, Yonatan and
Hupkes, Dieuwke",
booktitle = "Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP",
month = aug,
year = "2019",
address = "Florence, Italy",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/W19-4825",
doi = "10.18653/v1/W19-4825",
pages = "241--253",
abstract = "How and to what extent does BERT encode syntactically-sensitive hierarchical information or positionally-sensitive linear information? Recent work has shown that contextual representations like BERT perform well on tasks that require sensitivity to linguistic structure. We present here two studies which aim to provide a better understanding of the nature of BERT{'}s representations. The first of these focuses on the identification of structurally-defined elements using diagnostic classifiers, while the second explores BERT{'}s representation of subject-verb agreement and anaphor-antecedent dependencies through a quantitative assessment of self-attention vectors. In both cases, we find that BERT encodes positional information about word tokens well on its lower layers, but switches to a hierarchically-oriented encoding on higher layers. We conclude then that BERT{'}s representations do indeed model linguistically relevant aspects of hierarchical structure, though they do not appear to show the sharp sensitivity to hierarchical structure that is found in human processing of reflexive anaphora.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lin-etal-2019-open">
<titleInfo>
<title>Open Sesame: Getting inside BERT’s Linguistic Knowledge</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yongjie</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="given">Chern</namePart>
<namePart type="family">Tan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Robert</namePart>
<namePart type="family">Frank</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2019-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tal</namePart>
<namePart type="family">Linzen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Grzegorz</namePart>
<namePart type="family">Chrupała</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yonatan</namePart>
<namePart type="family">Belinkov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dieuwke</namePart>
<namePart type="family">Hupkes</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Florence, Italy</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>How and to what extent does BERT encode syntactically-sensitive hierarchical information or positionally-sensitive linear information? Recent work has shown that contextual representations like BERT perform well on tasks that require sensitivity to linguistic structure. We present here two studies which aim to provide a better understanding of the nature of BERT’s representations. The first of these focuses on the identification of structurally-defined elements using diagnostic classifiers, while the second explores BERT’s representation of subject-verb agreement and anaphor-antecedent dependencies through a quantitative assessment of self-attention vectors. In both cases, we find that BERT encodes positional information about word tokens well on its lower layers, but switches to a hierarchically-oriented encoding on higher layers. We conclude then that BERT’s representations do indeed model linguistically relevant aspects of hierarchical structure, though they do not appear to show the sharp sensitivity to hierarchical structure that is found in human processing of reflexive anaphora.</abstract>
<identifier type="citekey">lin-etal-2019-open</identifier>
<identifier type="doi">10.18653/v1/W19-4825</identifier>
<location>
<url>https://aclanthology.org/W19-4825</url>
</location>
<part>
<date>2019-08</date>
<extent unit="page">
<start>241</start>
<end>253</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Open Sesame: Getting inside BERT’s Linguistic Knowledge
%A Lin, Yongjie
%A Tan, Yi Chern
%A Frank, Robert
%Y Linzen, Tal
%Y Chrupała, Grzegorz
%Y Belinkov, Yonatan
%Y Hupkes, Dieuwke
%S Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP
%D 2019
%8 August
%I Association for Computational Linguistics
%C Florence, Italy
%F lin-etal-2019-open
%X How and to what extent does BERT encode syntactically-sensitive hierarchical information or positionally-sensitive linear information? Recent work has shown that contextual representations like BERT perform well on tasks that require sensitivity to linguistic structure. We present here two studies which aim to provide a better understanding of the nature of BERT’s representations. The first of these focuses on the identification of structurally-defined elements using diagnostic classifiers, while the second explores BERT’s representation of subject-verb agreement and anaphor-antecedent dependencies through a quantitative assessment of self-attention vectors. In both cases, we find that BERT encodes positional information about word tokens well on its lower layers, but switches to a hierarchically-oriented encoding on higher layers. We conclude then that BERT’s representations do indeed model linguistically relevant aspects of hierarchical structure, though they do not appear to show the sharp sensitivity to hierarchical structure that is found in human processing of reflexive anaphora.
%R 10.18653/v1/W19-4825
%U https://aclanthology.org/W19-4825
%U https://doi.org/10.18653/v1/W19-4825
%P 241-253
Markdown (Informal)
[Open Sesame: Getting inside BERT’s Linguistic Knowledge](https://aclanthology.org/W19-4825) (Lin et al., BlackboxNLP 2019)
ACL
- Yongjie Lin, Yi Chern Tan, and Robert Frank. 2019. Open Sesame: Getting inside BERT’s Linguistic Knowledge. In Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 241–253, Florence, Italy. Association for Computational Linguistics.