Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer

Scarcity of parallel data causes formality style transfer models to have scarce success in preserving content. We show that fine-tuning pre-trained language (GPT-2) and sequence-to-sequence (BART) models boosts content preservation, and that this is possible even with limited amounts of parallel dat...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Lai, Huiyuan, Toral, Antonio, Nissim, Malvina
Format:	Artikel
Sprache:	eng
Schlagworte:	Computer Science - Computation and Language
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	Lai, Huiyuan Toral, Antonio Nissim, Malvina
description	Scarcity of parallel data causes formality style transfer models to have scarce success in preserving content. We show that fine-tuning pre-trained language (GPT-2) and sequence-to-sequence (BART) models boosts content preservation, and that this is possible even with limited amounts of parallel data. Augmenting these models with rewards that target style and content -- the two core aspects of the task -- we achieve a new state-of-the-art.
doi_str_mv	10.48550/arxiv.2105.06947
format	Article
fullrecord	<record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2105_06947</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2105_06947</sourcerecordid><originalsourceid>FETCH-LOGICAL-a677-acd7b02414adbba48a5d9421cece41da44aeea65cc8bf4b255a2e823bf1900513</originalsourceid><addsrcrecordid>eNotz8lOwzAUhWFvWKDCA7DCPECC7VxnWJaK0kplUPE-urZvwCJD5ZTSvD1QWJ3NryN9jF1JkUKptbjFeAyHVEmhU5FXUJyzJ_OO_Qefhk9-N9-aG76lL4w-9G_8JVJiIoaePH8cPLUjX3e7OBxo5MshdtiG_cRf91NL_Kfrx4biBTtrsB3p8n9nzCzvzWKVbJ4f1ov5JsG8KBJ0vrBCgQT01iKUqH0FSjpyBNIjABJhrp0rbQNWaY2KSpXZRlZCaJnN2PXf7QlU72LoME71L6w-wbJvTmJJFQ</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer</title><source>arXiv.org</source><creator>Lai, Huiyuan ; Toral, Antonio ; Nissim, Malvina</creator><creatorcontrib>Lai, Huiyuan ; Toral, Antonio ; Nissim, Malvina</creatorcontrib><description>Scarcity of parallel data causes formality style transfer models to have scarce success in preserving content. We show that fine-tuning pre-trained language (GPT-2) and sequence-to-sequence (BART) models boosts content preservation, and that this is possible even with limited amounts of parallel data. Augmenting these models with rewards that target style and content -- the two core aspects of the task -- we achieve a new state-of-the-art.</description><identifier>DOI: 10.48550/arxiv.2105.06947</identifier><language>eng</language><subject>Computer Science - Computation and Language</subject><creationdate>2021-05</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</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/2105.06947$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2105.06947$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Lai, Huiyuan</creatorcontrib><creatorcontrib>Toral, Antonio</creatorcontrib><creatorcontrib>Nissim, Malvina</creatorcontrib><title>Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer</title><description>Scarcity of parallel data causes formality style transfer models to have scarce success in preserving content. We show that fine-tuning pre-trained language (GPT-2) and sequence-to-sequence (BART) models boosts content preservation, and that this is possible even with limited amounts of parallel data. Augmenting these models with rewards that target style and content -- the two core aspects of the task -- we achieve a new state-of-the-art.</description><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz8lOwzAUhWFvWKDCA7DCPECC7VxnWJaK0kplUPE-urZvwCJD5ZTSvD1QWJ3NryN9jF1JkUKptbjFeAyHVEmhU5FXUJyzJ_OO_Qefhk9-N9-aG76lL4w-9G_8JVJiIoaePH8cPLUjX3e7OBxo5MshdtiG_cRf91NL_Kfrx4biBTtrsB3p8n9nzCzvzWKVbJ4f1ov5JsG8KBJ0vrBCgQT01iKUqH0FSjpyBNIjABJhrp0rbQNWaY2KSpXZRlZCaJnN2PXf7QlU72LoME71L6w-wbJvTmJJFQ</recordid><startdate>20210514</startdate><enddate>20210514</enddate><creator>Lai, Huiyuan</creator><creator>Toral, Antonio</creator><creator>Nissim, Malvina</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20210514</creationdate><title>Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer</title><author>Lai, Huiyuan ; Toral, Antonio ; Nissim, Malvina</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a677-acd7b02414adbba48a5d9421cece41da44aeea65cc8bf4b255a2e823bf1900513</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Lai, Huiyuan</creatorcontrib><creatorcontrib>Toral, Antonio</creatorcontrib><creatorcontrib>Nissim, Malvina</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Lai, Huiyuan</au><au>Toral, Antonio</au><au>Nissim, Malvina</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer</atitle><date>2021-05-14</date><risdate>2021</risdate><abstract>Scarcity of parallel data causes formality style transfer models to have scarce success in preserving content. We show that fine-tuning pre-trained language (GPT-2) and sequence-to-sequence (BART) models boosts content preservation, and that this is possible even with limited amounts of parallel data. Augmenting these models with rewards that target style and content -- the two core aspects of the task -- we achieve a new state-of-the-art.</abstract><doi>10.48550/arxiv.2105.06947</doi><oa>free_for_read</oa></addata></record>
fulltext	fulltext_linktorsrc
identifier	DOI: 10.48550/arxiv.2105.06947
ispartof
issn
language	eng
recordid	cdi_arxiv_primary_2105_06947
source	arXiv.org
subjects	Computer Science - Computation and Language
title	Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer
url	https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-11T16%3A46%3A23IST&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=Thank%20you%20BART!%20Rewarding%20Pre-Trained%20Models%20Improves%20Formality%20Style%20Transfer&rft.au=Lai,%20Huiyuan&rft.date=2021-05-14&rft_id=info:doi/10.48550/arxiv.2105.06947&rft_dat=%3Carxiv_GOX%3E2105_06947%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