@inproceedings{yang-etal-2024-masking,
title = "Masking Latent Gender Knowledge for Debiasing Image Captioning",
author = "Yang, Fan and
Ghosh, Shalini and
Barut, Emre and
Qin, Kechen and
Wanigasekara, Prashan and
Su, Chengwei and
Ruan, Weitong and
Gupta, Rahul",
editor = "Ovalle, Anaelia and
Chang, Kai-Wei and
Cao, Yang Trista and
Mehrabi, Ninareh and
Zhao, Jieyu and
Galstyan, Aram and
Dhamala, Jwala and
Kumar, Anoop and
Gupta, Rahul",
booktitle = "Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.trustnlp-1.19/",
doi = "10.18653/v1/2024.trustnlp-1.19",
pages = "227--238",
abstract = "Large language models incorporate world knowledge and present breakthrough performances on zero-shot learning. However, these models capture societal bias (e.g., gender or racial bias) due to bias during the training process which raises ethical concerns or can even be potentially harmful. The issue is more pronounced in multi-modal settings, such as image captioning, as images can also add onto biases (e.g., due to historical non-equal representation of genders in different occupations). In this study, we investigate the removal of potentially problematic knowledge from multi-modal models used for image captioning. We relax the gender bias issue in captioning models by degenderizing generated captions through the use of a simple linear mask, trained via adversarial training. Our proposal makes no assumption on the architecture of the model and freezes the model weights during the procedure, which also enables the mask to be turned off. We conduct experiments on COCO caption datasets using our masking solution. The results suggest that the proposed mechanism can effectively mask the targeted biased knowledge, by replacing more than 99{\%} gender words with neutral ones, and maintain a comparable captioning quality performance with minimal (e.g., -1.4 on BLEU4 and ROUGE) impact to accuracy metrics."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yang-etal-2024-masking">
<titleInfo>
<title>Masking Latent Gender Knowledge for Debiasing Image Captioning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Fan</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shalini</namePart>
<namePart type="family">Ghosh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Emre</namePart>
<namePart type="family">Barut</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kechen</namePart>
<namePart type="family">Qin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Prashan</namePart>
<namePart type="family">Wanigasekara</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chengwei</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weitong</namePart>
<namePart type="family">Ruan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-06</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Anaelia</namePart>
<namePart type="family">Ovalle</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai-Wei</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="given">Trista</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ninareh</namePart>
<namePart type="family">Mehrabi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jieyu</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aram</namePart>
<namePart type="family">Galstyan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jwala</namePart>
<namePart type="family">Dhamala</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anoop</namePart>
<namePart type="family">Kumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Mexico City, Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large language models incorporate world knowledge and present breakthrough performances on zero-shot learning. However, these models capture societal bias (e.g., gender or racial bias) due to bias during the training process which raises ethical concerns or can even be potentially harmful. The issue is more pronounced in multi-modal settings, such as image captioning, as images can also add onto biases (e.g., due to historical non-equal representation of genders in different occupations). In this study, we investigate the removal of potentially problematic knowledge from multi-modal models used for image captioning. We relax the gender bias issue in captioning models by degenderizing generated captions through the use of a simple linear mask, trained via adversarial training. Our proposal makes no assumption on the architecture of the model and freezes the model weights during the procedure, which also enables the mask to be turned off. We conduct experiments on COCO caption datasets using our masking solution. The results suggest that the proposed mechanism can effectively mask the targeted biased knowledge, by replacing more than 99% gender words with neutral ones, and maintain a comparable captioning quality performance with minimal (e.g., -1.4 on BLEU4 and ROUGE) impact to accuracy metrics.</abstract>
<identifier type="citekey">yang-etal-2024-masking</identifier>
<identifier type="doi">10.18653/v1/2024.trustnlp-1.19</identifier>
<location>
<url>https://aclanthology.org/2024.trustnlp-1.19/</url>
</location>
<part>
<date>2024-06</date>
<extent unit="page">
<start>227</start>
<end>238</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Masking Latent Gender Knowledge for Debiasing Image Captioning
%A Yang, Fan
%A Ghosh, Shalini
%A Barut, Emre
%A Qin, Kechen
%A Wanigasekara, Prashan
%A Su, Chengwei
%A Ruan, Weitong
%A Gupta, Rahul
%Y Ovalle, Anaelia
%Y Chang, Kai-Wei
%Y Cao, Yang Trista
%Y Mehrabi, Ninareh
%Y Zhao, Jieyu
%Y Galstyan, Aram
%Y Dhamala, Jwala
%Y Kumar, Anoop
%Y Gupta, Rahul
%S Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024)
%D 2024
%8 June
%I Association for Computational Linguistics
%C Mexico City, Mexico
%F yang-etal-2024-masking
%X Large language models incorporate world knowledge and present breakthrough performances on zero-shot learning. However, these models capture societal bias (e.g., gender or racial bias) due to bias during the training process which raises ethical concerns or can even be potentially harmful. The issue is more pronounced in multi-modal settings, such as image captioning, as images can also add onto biases (e.g., due to historical non-equal representation of genders in different occupations). In this study, we investigate the removal of potentially problematic knowledge from multi-modal models used for image captioning. We relax the gender bias issue in captioning models by degenderizing generated captions through the use of a simple linear mask, trained via adversarial training. Our proposal makes no assumption on the architecture of the model and freezes the model weights during the procedure, which also enables the mask to be turned off. We conduct experiments on COCO caption datasets using our masking solution. The results suggest that the proposed mechanism can effectively mask the targeted biased knowledge, by replacing more than 99% gender words with neutral ones, and maintain a comparable captioning quality performance with minimal (e.g., -1.4 on BLEU4 and ROUGE) impact to accuracy metrics.
%R 10.18653/v1/2024.trustnlp-1.19
%U https://aclanthology.org/2024.trustnlp-1.19/
%U https://doi.org/10.18653/v1/2024.trustnlp-1.19
%P 227-238
Markdown (Informal)
[Masking Latent Gender Knowledge for Debiasing Image Captioning](https://aclanthology.org/2024.trustnlp-1.19/) (Yang et al., TrustNLP 2024)
ACL
- Fan Yang, Shalini Ghosh, Emre Barut, Kechen Qin, Prashan Wanigasekara, Chengwei Su, Weitong Ruan, and Rahul Gupta. 2024. Masking Latent Gender Knowledge for Debiasing Image Captioning. In Proceedings of the 4th Workshop on Trustworthy Natural Language Processing (TrustNLP 2024), pages 227–238, Mexico City, Mexico. Association for Computational Linguistics.