@inproceedings{wegge-klinger-2024-topic,
title = "Topic Bias in Emotion Classification",
author = "Wegge, Maximilian and
Klinger, Roman",
editor = {van der Goot, Rob and
Bak, JinYeong and
M{\"u}ller-Eberstein, Max and
Xu, Wei and
Ritter, Alan and
Baldwin, Tim},
booktitle = "Proceedings of the Ninth Workshop on Noisy and User-generated Text (W-NUT 2024)",
month = mar,
year = "2024",
address = "San {\.G}iljan, Malta",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.wnut-1.9",
pages = "89--103",
abstract = "Emotion corpora are typically sampled based on keyword/hashtag search or by asking study participants to generate textual instances. In any case, these corpora are not uniform samples representing the entirety of a domain. We hypothesize that this practice of data acquision leads to unrealistic correlations between overrepresented topics in these corpora that harm the generalizability of models. Such topic bias could lead to wrong predictions for instances like {``}I organized the service for my aunt{'}s funeral.{''} when funeral events are overpresented for instances labeled with sadness, despite the emotion of pride being more appropriate here. In this paper, we study this topic bias both from the data and the modeling perspective. We first label a set of emotion corpora automatically via topic modeling and show that emotions in fact correlate with specific topics. Further, we see that emotion classifiers are confounded by such topics. Finally, we show that the established debiasing method of adversarial correction via gradient reversal mitigates the issue. Our work points out issues with existing emotion corpora and that more representative resources are required for fair evaluation of models predicting affective concepts from text.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wegge-klinger-2024-topic">
<titleInfo>
<title>Topic Bias in Emotion Classification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maximilian</namePart>
<namePart type="family">Wegge</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roman</namePart>
<namePart type="family">Klinger</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-03</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Ninth Workshop on Noisy and User-generated Text (W-NUT 2024)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rob</namePart>
<namePart type="family">van der Goot</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">JinYeong</namePart>
<namePart type="family">Bak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Max</namePart>
<namePart type="family">Müller-Eberstein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tim</namePart>
<namePart type="family">Baldwin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Ġiljan, Malta</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Emotion corpora are typically sampled based on keyword/hashtag search or by asking study participants to generate textual instances. In any case, these corpora are not uniform samples representing the entirety of a domain. We hypothesize that this practice of data acquision leads to unrealistic correlations between overrepresented topics in these corpora that harm the generalizability of models. Such topic bias could lead to wrong predictions for instances like “I organized the service for my aunt’s funeral.” when funeral events are overpresented for instances labeled with sadness, despite the emotion of pride being more appropriate here. In this paper, we study this topic bias both from the data and the modeling perspective. We first label a set of emotion corpora automatically via topic modeling and show that emotions in fact correlate with specific topics. Further, we see that emotion classifiers are confounded by such topics. Finally, we show that the established debiasing method of adversarial correction via gradient reversal mitigates the issue. Our work points out issues with existing emotion corpora and that more representative resources are required for fair evaluation of models predicting affective concepts from text.</abstract>
<identifier type="citekey">wegge-klinger-2024-topic</identifier>
<location>
<url>https://aclanthology.org/2024.wnut-1.9</url>
</location>
<part>
<date>2024-03</date>
<extent unit="page">
<start>89</start>
<end>103</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Topic Bias in Emotion Classification
%A Wegge, Maximilian
%A Klinger, Roman
%Y van der Goot, Rob
%Y Bak, JinYeong
%Y Müller-Eberstein, Max
%Y Xu, Wei
%Y Ritter, Alan
%Y Baldwin, Tim
%S Proceedings of the Ninth Workshop on Noisy and User-generated Text (W-NUT 2024)
%D 2024
%8 March
%I Association for Computational Linguistics
%C San Ġiljan, Malta
%F wegge-klinger-2024-topic
%X Emotion corpora are typically sampled based on keyword/hashtag search or by asking study participants to generate textual instances. In any case, these corpora are not uniform samples representing the entirety of a domain. We hypothesize that this practice of data acquision leads to unrealistic correlations between overrepresented topics in these corpora that harm the generalizability of models. Such topic bias could lead to wrong predictions for instances like “I organized the service for my aunt’s funeral.” when funeral events are overpresented for instances labeled with sadness, despite the emotion of pride being more appropriate here. In this paper, we study this topic bias both from the data and the modeling perspective. We first label a set of emotion corpora automatically via topic modeling and show that emotions in fact correlate with specific topics. Further, we see that emotion classifiers are confounded by such topics. Finally, we show that the established debiasing method of adversarial correction via gradient reversal mitigates the issue. Our work points out issues with existing emotion corpora and that more representative resources are required for fair evaluation of models predicting affective concepts from text.
%U https://aclanthology.org/2024.wnut-1.9
%P 89-103
Markdown (Informal)
[Topic Bias in Emotion Classification](https://aclanthology.org/2024.wnut-1.9) (Wegge & Klinger, WNUT-WS 2024)
ACL
- Maximilian Wegge and Roman Klinger. 2024. Topic Bias in Emotion Classification. In Proceedings of the Ninth Workshop on Noisy and User-generated Text (W-NUT 2024), pages 89–103, San Ġiljan, Malta. Association for Computational Linguistics.