@inproceedings{arnold-etal-2023-disentangling,
title = "Disentangling the Linguistic Competence of Privacy-Preserving {BERT}",
author = "Arnold, Stefan and
Kemmerzell, Nils and
Schreiner, Annika",
editor = "Belinkov, Yonatan and
Hao, Sophie and
Jumelet, Jaap and
Kim, Najoung and
McCarthy, Arya and
Mohebbi, Hosein",
booktitle = "Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.blackboxnlp-1.5",
doi = "10.18653/v1/2023.blackboxnlp-1.5",
pages = "65--75",
abstract = "Differential Privacy (DP) has been tailored to address the unique challenges of text-to-text privatization. However, text-to-text privatization is known for degrading the performance of language models when trained on perturbed text. Employing a series of interpretation techniques on the internal representations extracted from BERT trained on perturbed pre-text, we intend to disentangle at the linguistic level the distortion induced by differential privacy. Experimental results from a representational similarity analysis indicate that the overall similarity of internal representations is substantially reduced. Using probing tasks to unpack this dissimilarity, we find evidence that text-to-text privatization affects the linguistic competence across several formalisms, encoding localized properties of words while falling short at encoding the contextual relationships between spans of words.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="arnold-etal-2023-disentangling">
<titleInfo>
<title>Disentangling the Linguistic Competence of Privacy-Preserving BERT</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stefan</namePart>
<namePart type="family">Arnold</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nils</namePart>
<namePart type="family">Kemmerzell</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Annika</namePart>
<namePart type="family">Schreiner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP</title>
</titleInfo>
<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">Sophie</namePart>
<namePart type="family">Hao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jaap</namePart>
<namePart type="family">Jumelet</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Najoung</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arya</namePart>
<namePart type="family">McCarthy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hosein</namePart>
<namePart type="family">Mohebbi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Singapore</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Differential Privacy (DP) has been tailored to address the unique challenges of text-to-text privatization. However, text-to-text privatization is known for degrading the performance of language models when trained on perturbed text. Employing a series of interpretation techniques on the internal representations extracted from BERT trained on perturbed pre-text, we intend to disentangle at the linguistic level the distortion induced by differential privacy. Experimental results from a representational similarity analysis indicate that the overall similarity of internal representations is substantially reduced. Using probing tasks to unpack this dissimilarity, we find evidence that text-to-text privatization affects the linguistic competence across several formalisms, encoding localized properties of words while falling short at encoding the contextual relationships between spans of words.</abstract>
<identifier type="citekey">arnold-etal-2023-disentangling</identifier>
<identifier type="doi">10.18653/v1/2023.blackboxnlp-1.5</identifier>
<location>
<url>https://aclanthology.org/2023.blackboxnlp-1.5</url>
</location>
<part>
<date>2023-12</date>
<extent unit="page">
<start>65</start>
<end>75</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Disentangling the Linguistic Competence of Privacy-Preserving BERT
%A Arnold, Stefan
%A Kemmerzell, Nils
%A Schreiner, Annika
%Y Belinkov, Yonatan
%Y Hao, Sophie
%Y Jumelet, Jaap
%Y Kim, Najoung
%Y McCarthy, Arya
%Y Mohebbi, Hosein
%S Proceedings of the 6th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP
%D 2023
%8 December
%I Association for Computational Linguistics
%C Singapore
%F arnold-etal-2023-disentangling
%X Differential Privacy (DP) has been tailored to address the unique challenges of text-to-text privatization. However, text-to-text privatization is known for degrading the performance of language models when trained on perturbed text. Employing a series of interpretation techniques on the internal representations extracted from BERT trained on perturbed pre-text, we intend to disentangle at the linguistic level the distortion induced by differential privacy. Experimental results from a representational similarity analysis indicate that the overall similarity of internal representations is substantially reduced. Using probing tasks to unpack this dissimilarity, we find evidence that text-to-text privatization affects the linguistic competence across several formalisms, encoding localized properties of words while falling short at encoding the contextual relationships between spans of words.
%R 10.18653/v1/2023.blackboxnlp-1.5
%U https://aclanthology.org/2023.blackboxnlp-1.5
%U https://doi.org/10.18653/v1/2023.blackboxnlp-1.5
%P 65-75
Markdown (Informal)
[Disentangling the Linguistic Competence of Privacy-Preserving BERT](https://aclanthology.org/2023.blackboxnlp-1.5) (Arnold et al., BlackboxNLP-WS 2023)
ACL