@article{yin-etal-2024-source-free,
title = "Source-Free Domain Adaptation for Question Answering with Masked Self-training",
author = "Yin, Maxwell J. and
Wang, Boyu and
Dong, Yue and
Ling, Charles",
journal = "Transactions of the Association for Computational Linguistics",
volume = "12",
year = "2024",
address = "Cambridge, MA",
publisher = "MIT Press",
url = "https://aclanthology.org/2024.tacl-1.40",
doi = "10.1162/tacl_a_00669",
pages = "721--737",
abstract = "Previous unsupervised domain adaptation (UDA) methods for question answering (QA) require access to source domain data while fine-tuning the model for the target domain. Source domain data may, however, contain sensitive information and should be protected. In this study, we investigate a more challenging setting, source-free UDA, in which we have only the pretrained source model and target domain data, without access to source domain data. We propose a novel self-training approach to QA models that integrates a specially designed mask module for domain adaptation. The mask is auto-adjusted to extract key domain knowledge when trained on the source domain. To maintain previously learned domain knowledge, certain mask weights are frozen during adaptation, while other weights are adjusted to mitigate domain shifts with pseudo-labeled samples generated in the target domain. Our empirical results on four benchmark datasets suggest that our approach significantly enhances the performance of pretrained QA models on the target domain, and even outperforms models that have access to the source data during adaptation.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yin-etal-2024-source-free">
<titleInfo>
<title>Source-Free Domain Adaptation for Question Answering with Masked Self-training</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maxwell</namePart>
<namePart type="given">J</namePart>
<namePart type="family">Yin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Boyu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Dong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Charles</namePart>
<namePart type="family">Ling</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<genre authority="bibutilsgt">journal article</genre>
<relatedItem type="host">
<titleInfo>
<title>Transactions of the Association for Computational Linguistics</title>
</titleInfo>
<originInfo>
<issuance>continuing</issuance>
<publisher>MIT Press</publisher>
<place>
<placeTerm type="text">Cambridge, MA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">periodical</genre>
<genre authority="bibutilsgt">academic journal</genre>
</relatedItem>
<abstract>Previous unsupervised domain adaptation (UDA) methods for question answering (QA) require access to source domain data while fine-tuning the model for the target domain. Source domain data may, however, contain sensitive information and should be protected. In this study, we investigate a more challenging setting, source-free UDA, in which we have only the pretrained source model and target domain data, without access to source domain data. We propose a novel self-training approach to QA models that integrates a specially designed mask module for domain adaptation. The mask is auto-adjusted to extract key domain knowledge when trained on the source domain. To maintain previously learned domain knowledge, certain mask weights are frozen during adaptation, while other weights are adjusted to mitigate domain shifts with pseudo-labeled samples generated in the target domain. Our empirical results on four benchmark datasets suggest that our approach significantly enhances the performance of pretrained QA models on the target domain, and even outperforms models that have access to the source data during adaptation.</abstract>
<identifier type="citekey">yin-etal-2024-source-free</identifier>
<identifier type="doi">10.1162/tacl_a_00669</identifier>
<location>
<url>https://aclanthology.org/2024.tacl-1.40</url>
</location>
<part>
<date>2024</date>
<detail type="volume"><number>12</number></detail>
<extent unit="page">
<start>721</start>
<end>737</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Journal Article
%T Source-Free Domain Adaptation for Question Answering with Masked Self-training
%A Yin, Maxwell J.
%A Wang, Boyu
%A Dong, Yue
%A Ling, Charles
%J Transactions of the Association for Computational Linguistics
%D 2024
%V 12
%I MIT Press
%C Cambridge, MA
%F yin-etal-2024-source-free
%X Previous unsupervised domain adaptation (UDA) methods for question answering (QA) require access to source domain data while fine-tuning the model for the target domain. Source domain data may, however, contain sensitive information and should be protected. In this study, we investigate a more challenging setting, source-free UDA, in which we have only the pretrained source model and target domain data, without access to source domain data. We propose a novel self-training approach to QA models that integrates a specially designed mask module for domain adaptation. The mask is auto-adjusted to extract key domain knowledge when trained on the source domain. To maintain previously learned domain knowledge, certain mask weights are frozen during adaptation, while other weights are adjusted to mitigate domain shifts with pseudo-labeled samples generated in the target domain. Our empirical results on four benchmark datasets suggest that our approach significantly enhances the performance of pretrained QA models on the target domain, and even outperforms models that have access to the source data during adaptation.
%R 10.1162/tacl_a_00669
%U https://aclanthology.org/2024.tacl-1.40
%U https://doi.org/10.1162/tacl_a_00669
%P 721-737
Markdown (Informal)
[Source-Free Domain Adaptation for Question Answering with Masked Self-training](https://aclanthology.org/2024.tacl-1.40) (Yin et al., TACL 2024)
ACL