Build a Robust QA System with Transformer-based Mixture of Experts
In this paper, we aim to build a robust question answering system that can adapt to out-of-domain datasets. A single network may overfit to the superficial correlation in the training distribution, but with a meaningful number of expert sub-networks, a gating network that selects a sparse combinatio...
Gespeichert in:
Hauptverfasser: | , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | |
container_volume | |
creator | Zhou, Yu Qing Liu, Xixuan Julie Dong, Yuanzhe |
description | In this paper, we aim to build a robust question answering system that can
adapt to out-of-domain datasets. A single network may overfit to the
superficial correlation in the training distribution, but with a meaningful
number of expert sub-networks, a gating network that selects a sparse
combination of experts for each input, and careful balance on the importance of
expert sub-networks, the Mixture-of-Experts (MoE) model allows us to train a
multi-task learner that can be generalized to out-of-domain datasets. We also
explore the possibility of bringing the MoE layers up to the middle of the
DistilBERT and replacing the dense feed-forward network with a
sparsely-activated switch FFN layers, similar to the Switch Transformer
architecture, which simplifies the MoE routing algorithm with reduced
communication and computational costs. In addition to model architectures, we
explore techniques of data augmentation including Easy Data Augmentation (EDA)
and back translation, to create more meaningful variance among the small
out-of-domain training data, therefore boosting the performance and robustness
of our models. In this paper, we show that our combination of best architecture
and data augmentation techniques achieves a 53.477 F1 score in the
out-of-domain evaluation, which is a 9.52% performance gain over the baseline.
On the final test set, we reported a higher 59.506 F1 and 41.651 EM. We
successfully demonstrate the effectiveness of Mixture-of-Expert architecture in
a Robust QA task. |
doi_str_mv | 10.48550/arxiv.2204.09598 |
format | Article |
fullrecord | <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2204_09598</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2204_09598</sourcerecordid><originalsourceid>FETCH-LOGICAL-a678-7d8284ca7acc636c38323f9b1e16330f6db4ad201fdde345834b88ba480016763</originalsourceid><addsrcrecordid>eNotz71OwzAUQGEvDKjwAEz4BRLsXMe5Gduq_EhFCMgeXce2sNSQynYgfXtEYTrbkT7GbqQoFda1uKO4hK-yqoQqRVu3eMk2mzkcLCf-Npk5Zf665u-nlN3Iv0P-4F2kz-SnOLpYGErO8uew5Dk6Pnm-W44u5nTFLjwdkrv-74p197tu-1jsXx6etut9QbrBorFYoRqooWHQoAdAqMC3RjqpAYTX1iiylZDeWgeqRlAG0ZBCIaRuNKzY7d_2jOiPMYwUT_0vpj9j4AdIekOm</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Build a Robust QA System with Transformer-based Mixture of Experts</title><source>arXiv.org</source><creator>Zhou, Yu Qing ; Liu, Xixuan Julie ; Dong, Yuanzhe</creator><creatorcontrib>Zhou, Yu Qing ; Liu, Xixuan Julie ; Dong, Yuanzhe</creatorcontrib><description>In this paper, we aim to build a robust question answering system that can
adapt to out-of-domain datasets. A single network may overfit to the
superficial correlation in the training distribution, but with a meaningful
number of expert sub-networks, a gating network that selects a sparse
combination of experts for each input, and careful balance on the importance of
expert sub-networks, the Mixture-of-Experts (MoE) model allows us to train a
multi-task learner that can be generalized to out-of-domain datasets. We also
explore the possibility of bringing the MoE layers up to the middle of the
DistilBERT and replacing the dense feed-forward network with a
sparsely-activated switch FFN layers, similar to the Switch Transformer
architecture, which simplifies the MoE routing algorithm with reduced
communication and computational costs. In addition to model architectures, we
explore techniques of data augmentation including Easy Data Augmentation (EDA)
and back translation, to create more meaningful variance among the small
out-of-domain training data, therefore boosting the performance and robustness
of our models. In this paper, we show that our combination of best architecture
and data augmentation techniques achieves a 53.477 F1 score in the
out-of-domain evaluation, which is a 9.52% performance gain over the baseline.
On the final test set, we reported a higher 59.506 F1 and 41.651 EM. We
successfully demonstrate the effectiveness of Mixture-of-Expert architecture in
a Robust QA task.</description><identifier>DOI: 10.48550/arxiv.2204.09598</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computation and Language</subject><creationdate>2022-03</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2204.09598$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2204.09598$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Zhou, Yu Qing</creatorcontrib><creatorcontrib>Liu, Xixuan Julie</creatorcontrib><creatorcontrib>Dong, Yuanzhe</creatorcontrib><title>Build a Robust QA System with Transformer-based Mixture of Experts</title><description>In this paper, we aim to build a robust question answering system that can
adapt to out-of-domain datasets. A single network may overfit to the
superficial correlation in the training distribution, but with a meaningful
number of expert sub-networks, a gating network that selects a sparse
combination of experts for each input, and careful balance on the importance of
expert sub-networks, the Mixture-of-Experts (MoE) model allows us to train a
multi-task learner that can be generalized to out-of-domain datasets. We also
explore the possibility of bringing the MoE layers up to the middle of the
DistilBERT and replacing the dense feed-forward network with a
sparsely-activated switch FFN layers, similar to the Switch Transformer
architecture, which simplifies the MoE routing algorithm with reduced
communication and computational costs. In addition to model architectures, we
explore techniques of data augmentation including Easy Data Augmentation (EDA)
and back translation, to create more meaningful variance among the small
out-of-domain training data, therefore boosting the performance and robustness
of our models. In this paper, we show that our combination of best architecture
and data augmentation techniques achieves a 53.477 F1 score in the
out-of-domain evaluation, which is a 9.52% performance gain over the baseline.
On the final test set, we reported a higher 59.506 F1 and 41.651 EM. We
successfully demonstrate the effectiveness of Mixture-of-Expert architecture in
a Robust QA task.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz71OwzAUQGEvDKjwAEz4BRLsXMe5Gduq_EhFCMgeXce2sNSQynYgfXtEYTrbkT7GbqQoFda1uKO4hK-yqoQqRVu3eMk2mzkcLCf-Npk5Zf665u-nlN3Iv0P-4F2kz-SnOLpYGErO8uew5Dk6Pnm-W44u5nTFLjwdkrv-74p197tu-1jsXx6etut9QbrBorFYoRqooWHQoAdAqMC3RjqpAYTX1iiylZDeWgeqRlAG0ZBCIaRuNKzY7d_2jOiPMYwUT_0vpj9j4AdIekOm</recordid><startdate>20220319</startdate><enddate>20220319</enddate><creator>Zhou, Yu Qing</creator><creator>Liu, Xixuan Julie</creator><creator>Dong, Yuanzhe</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20220319</creationdate><title>Build a Robust QA System with Transformer-based Mixture of Experts</title><author>Zhou, Yu Qing ; Liu, Xixuan Julie ; Dong, Yuanzhe</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a678-7d8284ca7acc636c38323f9b1e16330f6db4ad201fdde345834b88ba480016763</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Zhou, Yu Qing</creatorcontrib><creatorcontrib>Liu, Xixuan Julie</creatorcontrib><creatorcontrib>Dong, Yuanzhe</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Zhou, Yu Qing</au><au>Liu, Xixuan Julie</au><au>Dong, Yuanzhe</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Build a Robust QA System with Transformer-based Mixture of Experts</atitle><date>2022-03-19</date><risdate>2022</risdate><abstract>In this paper, we aim to build a robust question answering system that can
adapt to out-of-domain datasets. A single network may overfit to the
superficial correlation in the training distribution, but with a meaningful
number of expert sub-networks, a gating network that selects a sparse
combination of experts for each input, and careful balance on the importance of
expert sub-networks, the Mixture-of-Experts (MoE) model allows us to train a
multi-task learner that can be generalized to out-of-domain datasets. We also
explore the possibility of bringing the MoE layers up to the middle of the
DistilBERT and replacing the dense feed-forward network with a
sparsely-activated switch FFN layers, similar to the Switch Transformer
architecture, which simplifies the MoE routing algorithm with reduced
communication and computational costs. In addition to model architectures, we
explore techniques of data augmentation including Easy Data Augmentation (EDA)
and back translation, to create more meaningful variance among the small
out-of-domain training data, therefore boosting the performance and robustness
of our models. In this paper, we show that our combination of best architecture
and data augmentation techniques achieves a 53.477 F1 score in the
out-of-domain evaluation, which is a 9.52% performance gain over the baseline.
On the final test set, we reported a higher 59.506 F1 and 41.651 EM. We
successfully demonstrate the effectiveness of Mixture-of-Expert architecture in
a Robust QA task.</abstract><doi>10.48550/arxiv.2204.09598</doi><oa>free_for_read</oa></addata></record> |
fulltext | fulltext_linktorsrc |
identifier | DOI: 10.48550/arxiv.2204.09598 |
ispartof | |
issn | |
language | eng |
recordid | cdi_arxiv_primary_2204_09598 |
source | arXiv.org |
subjects | Computer Science - Artificial Intelligence Computer Science - Computation and Language |
title | Build a Robust QA System with Transformer-based Mixture of Experts |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-03T07%3A38%3A01IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Build%20a%20Robust%20QA%20System%20with%20Transformer-based%20Mixture%20of%20Experts&rft.au=Zhou,%20Yu%20Qing&rft.date=2022-03-19&rft_id=info:doi/10.48550/arxiv.2204.09598&rft_dat=%3Carxiv_GOX%3E2204_09598%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true |