Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization

Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization

Token embeddings in multilingual BERT (m-BERT) contain both language and semantic information. We find that the representation of a language can be obtained by simply averaging the embeddings of the tokens of the language. Given this language representation, we control the output languages of multil...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2021-11
Hauptverfasser: Chi-Liang, Liu, Tsung-Yuan Hsu, Yung-Sung Chuang, Chung-Yi, Li, Hung-yi, Lee
Format: Artikel
Sprache:eng
Schlagworte:
Language
Multilingualism
Representations
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Chi-Liang, Liu
Tsung-Yuan Hsu
Yung-Sung Chuang
Chung-Yi, Li
Hung-yi, Lee
description Token embeddings in multilingual BERT (m-BERT) contain both language and semantic information. We find that the representation of a language can be obtained by simply averaging the embeddings of the tokens of the language. Given this language representation, we control the output languages of multilingual BERT by manipulating the token embeddings, thus achieving unsupervised token translation. We further propose a computationally cheap but effective approach to improve the cross-lingual ability of m-BERT based on this observation.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2453528599</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2453528599</sourcerecordid><originalsourceid>FETCH-proquest_journals_24535285993</originalsourceid><addsrcrecordid>eNqNjb0KwjAYRYMgWLTv8IFzoSaNtqul_kBdpJNLyZCW1JhP8-Pg01tBd6cL95zLnZCIMrZK8ozSGYmdG9I0pesN5ZxF5FIjXpXpoUMLpQ7SAXZQC9MH0UtQBk5Be6XVp9Cwrc4NeITj7W7xKaG06Fzyo3tppBVavYRXaBZk2gntZPzNOVnuqqY8JOP0MR75dsBgzYhamnHGac6Lgv1nvQHcTkJV</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2453528599</pqid></control><display><type>article</type><title>Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization</title><source>Free E- Journals</source><creator>Chi-Liang, Liu ; Tsung-Yuan Hsu ; Yung-Sung Chuang ; Chung-Yi, Li ; Hung-yi, Lee</creator><creatorcontrib>Chi-Liang, Liu ; Tsung-Yuan Hsu ; Yung-Sung Chuang ; Chung-Yi, Li ; Hung-yi, Lee</creatorcontrib><description>Token embeddings in multilingual BERT (m-BERT) contain both language and semantic information. We find that the representation of a language can be obtained by simply averaging the embeddings of the tokens of the language. Given this language representation, we control the output languages of multilingual BERT by manipulating the token embeddings, thus achieving unsupervised token translation. We further propose a computationally cheap but effective approach to improve the cross-lingual ability of m-BERT based on this observation.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Language ; Multilingualism ; Representations</subject><ispartof>arXiv.org, 2021-11</ispartof><rights>2021. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</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>776,780</link.rule.ids></links><search><creatorcontrib>Chi-Liang, Liu</creatorcontrib><creatorcontrib>Tsung-Yuan Hsu</creatorcontrib><creatorcontrib>Yung-Sung Chuang</creatorcontrib><creatorcontrib>Chung-Yi, Li</creatorcontrib><creatorcontrib>Hung-yi, Lee</creatorcontrib><title>Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization</title><title>arXiv.org</title><description>Token embeddings in multilingual BERT (m-BERT) contain both language and semantic information. We find that the representation of a language can be obtained by simply averaging the embeddings of the tokens of the language. Given this language representation, we control the output languages of multilingual BERT by manipulating the token embeddings, thus achieving unsupervised token translation. We further propose a computationally cheap but effective approach to improve the cross-lingual ability of m-BERT based on this observation.</description><subject>Language</subject><subject>Multilingualism</subject><subject>Representations</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><recordid>eNqNjb0KwjAYRYMgWLTv8IFzoSaNtqul_kBdpJNLyZCW1JhP8-Pg01tBd6cL95zLnZCIMrZK8ozSGYmdG9I0pesN5ZxF5FIjXpXpoUMLpQ7SAXZQC9MH0UtQBk5Be6XVp9Cwrc4NeITj7W7xKaG06Fzyo3tppBVavYRXaBZk2gntZPzNOVnuqqY8JOP0MR75dsBgzYhamnHGac6Lgv1nvQHcTkJV</recordid><startdate>20211101</startdate><enddate>20211101</enddate><creator>Chi-Liang, Liu</creator><creator>Tsung-Yuan Hsu</creator><creator>Yung-Sung Chuang</creator><creator>Chung-Yi, Li</creator><creator>Hung-yi, Lee</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20211101</creationdate><title>Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization</title><author>Chi-Liang, Liu ; Tsung-Yuan Hsu ; Yung-Sung Chuang ; Chung-Yi, Li ; Hung-yi, Lee</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_24535285993</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Language</topic><topic>Multilingualism</topic><topic>Representations</topic><toplevel>online_resources</toplevel><creatorcontrib>Chi-Liang, Liu</creatorcontrib><creatorcontrib>Tsung-Yuan Hsu</creatorcontrib><creatorcontrib>Yung-Sung Chuang</creatorcontrib><creatorcontrib>Chung-Yi, Li</creatorcontrib><creatorcontrib>Hung-yi, Lee</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Chi-Liang, Liu</au><au>Tsung-Yuan Hsu</au><au>Yung-Sung Chuang</au><au>Chung-Yi, Li</au><au>Hung-yi, Lee</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization</atitle><jtitle>arXiv.org</jtitle><date>2021-11-01</date><risdate>2021</risdate><eissn>2331-8422</eissn><abstract>Token embeddings in multilingual BERT (m-BERT) contain both language and semantic information. We find that the representation of a language can be obtained by simply averaging the embeddings of the tokens of the language. Given this language representation, we control the output languages of multilingual BERT by manipulating the token embeddings, thus achieving unsupervised token translation. We further propose a computationally cheap but effective approach to improve the cross-lingual ability of m-BERT based on this observation.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2021-11
issn 2331-8422
language eng
recordid cdi_proquest_journals_2453528599
source Free E- Journals
subjects Language
Multilingualism
Representations
title Looking for Clues of Language in Multilingual BERT to Improve Cross-lingual Generalization
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-03T04%3A57%3A09IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Looking%20for%20Clues%20of%20Language%20in%20Multilingual%20BERT%20to%20Improve%20Cross-lingual%20Generalization&rft.jtitle=arXiv.org&rft.au=Chi-Liang,%20Liu&rft.date=2021-11-01&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2453528599%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2453528599&rft_id=info:pmid/&rfr_iscdi=true