@inproceedings{dayanik-pado-2021-disentangling,
title = "Disentangling Document Topic and Author Gender in Multiple Languages: Lessons for Adversarial Debiasing",
author = "Dayanik, Erenay and
Pad{\'o}, Sebastian",
editor = "De Clercq, Orphee and
Balahur, Alexandra and
Sedoc, Joao and
Barriere, Valentin and
Tafreshi, Shabnam and
Buechel, Sven and
Hoste, Veronique",
booktitle = "Proceedings of the Eleventh Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis",
month = apr,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.wassa-1.6",
pages = "50--61",
abstract = "Text classification is a central tool in NLP. However, when the target classes are strongly correlated with other textual attributes, text classification models can pick up {``}wrong{''} features, leading to bad generalization and biases. In social media analysis, this problem surfaces for demographic user classes such as language, topic, or gender, which influence the generate text to a substantial extent. Adversarial training has been claimed to mitigate this problem, but thorough evaluation is missing. In this paper, we experiment with text classification of the correlated attributes of document topic and author gender, using a novel multilingual parallel corpus of TED talk transcripts. Our findings are: (a) individual classifiers for topic and author gender are indeed biased; (b) debiasing with adversarial training works for topic, but breaks down for author gender; (c) gender debiasing results differ across languages. We interpret the result in terms of feature space overlap, highlighting the role of linguistic surface realization of the target classes.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dayanik-pado-2021-disentangling">
<titleInfo>
<title>Disentangling Document Topic and Author Gender in Multiple Languages: Lessons for Adversarial Debiasing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Erenay</namePart>
<namePart type="family">Dayanik</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Padó</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Eleventh Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis</title>
</titleInfo>
<name type="personal">
<namePart type="given">Orphee</namePart>
<namePart type="family">De Clercq</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alexandra</namePart>
<namePart type="family">Balahur</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joao</namePart>
<namePart type="family">Sedoc</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Valentin</namePart>
<namePart type="family">Barriere</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shabnam</namePart>
<namePart type="family">Tafreshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sven</namePart>
<namePart type="family">Buechel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Veronique</namePart>
<namePart type="family">Hoste</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Text classification is a central tool in NLP. However, when the target classes are strongly correlated with other textual attributes, text classification models can pick up “wrong” features, leading to bad generalization and biases. In social media analysis, this problem surfaces for demographic user classes such as language, topic, or gender, which influence the generate text to a substantial extent. Adversarial training has been claimed to mitigate this problem, but thorough evaluation is missing. In this paper, we experiment with text classification of the correlated attributes of document topic and author gender, using a novel multilingual parallel corpus of TED talk transcripts. Our findings are: (a) individual classifiers for topic and author gender are indeed biased; (b) debiasing with adversarial training works for topic, but breaks down for author gender; (c) gender debiasing results differ across languages. We interpret the result in terms of feature space overlap, highlighting the role of linguistic surface realization of the target classes.</abstract>
<identifier type="citekey">dayanik-pado-2021-disentangling</identifier>
<location>
<url>https://aclanthology.org/2021.wassa-1.6</url>
</location>
<part>
<date>2021-04</date>
<extent unit="page">
<start>50</start>
<end>61</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Disentangling Document Topic and Author Gender in Multiple Languages: Lessons for Adversarial Debiasing
%A Dayanik, Erenay
%A Padó, Sebastian
%Y De Clercq, Orphee
%Y Balahur, Alexandra
%Y Sedoc, Joao
%Y Barriere, Valentin
%Y Tafreshi, Shabnam
%Y Buechel, Sven
%Y Hoste, Veronique
%S Proceedings of the Eleventh Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis
%D 2021
%8 April
%I Association for Computational Linguistics
%C Online
%F dayanik-pado-2021-disentangling
%X Text classification is a central tool in NLP. However, when the target classes are strongly correlated with other textual attributes, text classification models can pick up “wrong” features, leading to bad generalization and biases. In social media analysis, this problem surfaces for demographic user classes such as language, topic, or gender, which influence the generate text to a substantial extent. Adversarial training has been claimed to mitigate this problem, but thorough evaluation is missing. In this paper, we experiment with text classification of the correlated attributes of document topic and author gender, using a novel multilingual parallel corpus of TED talk transcripts. Our findings are: (a) individual classifiers for topic and author gender are indeed biased; (b) debiasing with adversarial training works for topic, but breaks down for author gender; (c) gender debiasing results differ across languages. We interpret the result in terms of feature space overlap, highlighting the role of linguistic surface realization of the target classes.
%U https://aclanthology.org/2021.wassa-1.6
%P 50-61
Markdown (Informal)
[Disentangling Document Topic and Author Gender in Multiple Languages: Lessons for Adversarial Debiasing](https://aclanthology.org/2021.wassa-1.6) (Dayanik & Padó, WASSA 2021)
ACL