A Supervised Word Alignment Method based on Cross-Language Span Prediction using Multilingual BERT

We present a novel supervised word alignment method based on cross-language span prediction. We first formalize a word alignment problem as a collection of independent predictions from a token in the source sentence to a span in the target sentence. As this is equivalent to a SQuAD v2.0 style questi...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Nagata, Masaaki, Katsuki, Chousa, Nishino, Masaaki
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 Nagata, Masaaki
Katsuki, Chousa
Nishino, Masaaki
description We present a novel supervised word alignment method based on cross-language span prediction. We first formalize a word alignment problem as a collection of independent predictions from a token in the source sentence to a span in the target sentence. As this is equivalent to a SQuAD v2.0 style question answering task, we then solve this problem by using multilingual BERT, which is fine-tuned on a manually created gold word alignment data. We greatly improved the word alignment accuracy by adding the context of the token to the question. In the experiments using five word alignment datasets among Chinese, Japanese, German, Romanian, French, and English, we show that the proposed method significantly outperformed previous supervised and unsupervised word alignment methods without using any bitexts for pretraining. For example, we achieved an F1 score of 86.7 for the Chinese-English data, which is 13.3 points higher than the previous state-of-the-art supervised methods.
doi_str_mv 10.48550/arxiv.2004.14516
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2004_14516</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2004_14516</sourcerecordid><originalsourceid>FETCH-LOGICAL-a676-8f90b115bd60506eaae031760b722dffdbbc88f0689bc1c3048f936a1f5d0c1a3</originalsourceid><addsrcrecordid>eNotj81OhDAUhbtxYUYfwJV9AfCW0sIskYw_CRONQ-KS3NIWmzCFFJjo2zuMrs7iO-ckHyF3DOI0FwIeMHy7U5wApDFLBZPXRBX0sIwmnNxkNP0cgqZF7zp_NH6mezN_DZoqXNngaRmGaYoq9N2CnaGHET19D0a7dnZnvEzOd3S_9LPr3drp6ePuo74hVxb7ydz-54bUT7u6fImqt-fXsqgilJmMcrsFxZhQWoIAaRANcJZJUFmSaGu1Um2eW5D5VrWs5ZCeF1wis0JDy5BvyP3f7UWyGYM7YvhpVtnmIst_ATHpT9E</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>A Supervised Word Alignment Method based on Cross-Language Span Prediction using Multilingual BERT</title><source>arXiv.org</source><creator>Nagata, Masaaki ; Katsuki, Chousa ; Nishino, Masaaki</creator><creatorcontrib>Nagata, Masaaki ; Katsuki, Chousa ; Nishino, Masaaki</creatorcontrib><description>We present a novel supervised word alignment method based on cross-language span prediction. We first formalize a word alignment problem as a collection of independent predictions from a token in the source sentence to a span in the target sentence. As this is equivalent to a SQuAD v2.0 style question answering task, we then solve this problem by using multilingual BERT, which is fine-tuned on a manually created gold word alignment data. We greatly improved the word alignment accuracy by adding the context of the token to the question. In the experiments using five word alignment datasets among Chinese, Japanese, German, Romanian, French, and English, we show that the proposed method significantly outperformed previous supervised and unsupervised word alignment methods without using any bitexts for pretraining. For example, we achieved an F1 score of 86.7 for the Chinese-English data, which is 13.3 points higher than the previous state-of-the-art supervised methods.</description><identifier>DOI: 10.48550/arxiv.2004.14516</identifier><language>eng</language><subject>Computer Science - Computation and Language</subject><creationdate>2020-04</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>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,781,886</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2004.14516$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2004.14516$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Nagata, Masaaki</creatorcontrib><creatorcontrib>Katsuki, Chousa</creatorcontrib><creatorcontrib>Nishino, Masaaki</creatorcontrib><title>A Supervised Word Alignment Method based on Cross-Language Span Prediction using Multilingual BERT</title><description>We present a novel supervised word alignment method based on cross-language span prediction. We first formalize a word alignment problem as a collection of independent predictions from a token in the source sentence to a span in the target sentence. As this is equivalent to a SQuAD v2.0 style question answering task, we then solve this problem by using multilingual BERT, which is fine-tuned on a manually created gold word alignment data. We greatly improved the word alignment accuracy by adding the context of the token to the question. In the experiments using five word alignment datasets among Chinese, Japanese, German, Romanian, French, and English, we show that the proposed method significantly outperformed previous supervised and unsupervised word alignment methods without using any bitexts for pretraining. For example, we achieved an F1 score of 86.7 for the Chinese-English data, which is 13.3 points higher than the previous state-of-the-art supervised methods.</description><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj81OhDAUhbtxYUYfwJV9AfCW0sIskYw_CRONQ-KS3NIWmzCFFJjo2zuMrs7iO-ckHyF3DOI0FwIeMHy7U5wApDFLBZPXRBX0sIwmnNxkNP0cgqZF7zp_NH6mezN_DZoqXNngaRmGaYoq9N2CnaGHET19D0a7dnZnvEzOd3S_9LPr3drp6ePuo74hVxb7ydz-54bUT7u6fImqt-fXsqgilJmMcrsFxZhQWoIAaRANcJZJUFmSaGu1Um2eW5D5VrWs5ZCeF1wis0JDy5BvyP3f7UWyGYM7YvhpVtnmIst_ATHpT9E</recordid><startdate>20200429</startdate><enddate>20200429</enddate><creator>Nagata, Masaaki</creator><creator>Katsuki, Chousa</creator><creator>Nishino, Masaaki</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20200429</creationdate><title>A Supervised Word Alignment Method based on Cross-Language Span Prediction using Multilingual BERT</title><author>Nagata, Masaaki ; Katsuki, Chousa ; Nishino, Masaaki</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a676-8f90b115bd60506eaae031760b722dffdbbc88f0689bc1c3048f936a1f5d0c1a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Nagata, Masaaki</creatorcontrib><creatorcontrib>Katsuki, Chousa</creatorcontrib><creatorcontrib>Nishino, Masaaki</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Nagata, Masaaki</au><au>Katsuki, Chousa</au><au>Nishino, Masaaki</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>A Supervised Word Alignment Method based on Cross-Language Span Prediction using Multilingual BERT</atitle><date>2020-04-29</date><risdate>2020</risdate><abstract>We present a novel supervised word alignment method based on cross-language span prediction. We first formalize a word alignment problem as a collection of independent predictions from a token in the source sentence to a span in the target sentence. As this is equivalent to a SQuAD v2.0 style question answering task, we then solve this problem by using multilingual BERT, which is fine-tuned on a manually created gold word alignment data. We greatly improved the word alignment accuracy by adding the context of the token to the question. In the experiments using five word alignment datasets among Chinese, Japanese, German, Romanian, French, and English, we show that the proposed method significantly outperformed previous supervised and unsupervised word alignment methods without using any bitexts for pretraining. For example, we achieved an F1 score of 86.7 for the Chinese-English data, which is 13.3 points higher than the previous state-of-the-art supervised methods.</abstract><doi>10.48550/arxiv.2004.14516</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2004.14516
ispartof
issn
language eng
recordid cdi_arxiv_primary_2004_14516
source arXiv.org
subjects Computer Science - Computation and Language
title A Supervised Word Alignment Method based on Cross-Language Span Prediction using Multilingual BERT
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-15T02%3A54%3A13IST&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=A%20Supervised%20Word%20Alignment%20Method%20based%20on%20Cross-Language%20Span%20Prediction%20using%20Multilingual%20BERT&rft.au=Nagata,%20Masaaki&rft.date=2020-04-29&rft_id=info:doi/10.48550/arxiv.2004.14516&rft_dat=%3Carxiv_GOX%3E2004_14516%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