Adversarial Training for Multilingual Acoustic Modeling

Multilingual training has been shown to improve acoustic modeling performance by sharing and transferring knowledge in modeling different languages. Knowledge sharing is usually achieved by using common lower-level layers for different languages in a deep neural network. Recently, the domain adversa...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Hu, Ke, Sak, Hasim, Liao, Hank
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 Hu, Ke
Sak, Hasim
Liao, Hank
description Multilingual training has been shown to improve acoustic modeling performance by sharing and transferring knowledge in modeling different languages. Knowledge sharing is usually achieved by using common lower-level layers for different languages in a deep neural network. Recently, the domain adversarial network was proposed to reduce domain mismatch of training data and learn domain-invariant features. It is thus worth exploring whether adversarial training can further promote knowledge sharing in multilingual models. In this work, we apply the domain adversarial network to encourage the shared layers of a multilingual model to learn language-invariant features. Bidirectional Long Short-Term Memory (LSTM) recurrent neural networks (RNN) are used as building blocks. We show that shared layers learned this way contain less language identification information and lead to better performance. In an automatic speech recognition task for seven languages, the resultant acoustic model improves the word error rate (WER) of the multilingual model by 4% relative on average, and the monolingual models by 10%.
doi_str_mv 10.48550/arxiv.1906.07093
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_1906_07093</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>1906_07093</sourcerecordid><originalsourceid>FETCH-LOGICAL-a673-f9cba7a5a9e4410cc7706128c687f5fd597369a856c0ffae135dfd6b0c910663</originalsourceid><addsrcrecordid>eNotj7uOwjAURN1QrIAP2Ir8QLLXOL6OyyhiYSUQBfTRxQ9kKSTIIQj-nsdSjWZGGs1h7JtDlhdSwg_FW7hmXANmoECLL6ZKe3WxpxioSfaRQhvaY-K7mGyG5hKapxueTWm6ob8Ek2w6617hhI08Nb2bfnTMdr-LfbVK19vlX1WuU0IlUq_NgRRJ0i7PORijFCCfFwYL5aW3UiuBmgqJBrwnx4W03uIBjOaAKMZs9r_6Pl6fYzhRvNcvgPoNIB5Jc0CW</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Adversarial Training for Multilingual Acoustic Modeling</title><source>arXiv.org</source><creator>Hu, Ke ; Sak, Hasim ; Liao, Hank</creator><creatorcontrib>Hu, Ke ; Sak, Hasim ; Liao, Hank</creatorcontrib><description>Multilingual training has been shown to improve acoustic modeling performance by sharing and transferring knowledge in modeling different languages. Knowledge sharing is usually achieved by using common lower-level layers for different languages in a deep neural network. Recently, the domain adversarial network was proposed to reduce domain mismatch of training data and learn domain-invariant features. It is thus worth exploring whether adversarial training can further promote knowledge sharing in multilingual models. In this work, we apply the domain adversarial network to encourage the shared layers of a multilingual model to learn language-invariant features. Bidirectional Long Short-Term Memory (LSTM) recurrent neural networks (RNN) are used as building blocks. We show that shared layers learned this way contain less language identification information and lead to better performance. In an automatic speech recognition task for seven languages, the resultant acoustic model improves the word error rate (WER) of the multilingual model by 4% relative on average, and the monolingual models by 10%.</description><identifier>DOI: 10.48550/arxiv.1906.07093</identifier><language>eng</language><subject>Computer Science - Computation and Language ; Computer Science - Learning ; Computer Science - Sound</subject><creationdate>2019-06</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/1906.07093$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.1906.07093$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Hu, Ke</creatorcontrib><creatorcontrib>Sak, Hasim</creatorcontrib><creatorcontrib>Liao, Hank</creatorcontrib><title>Adversarial Training for Multilingual Acoustic Modeling</title><description>Multilingual training has been shown to improve acoustic modeling performance by sharing and transferring knowledge in modeling different languages. Knowledge sharing is usually achieved by using common lower-level layers for different languages in a deep neural network. Recently, the domain adversarial network was proposed to reduce domain mismatch of training data and learn domain-invariant features. It is thus worth exploring whether adversarial training can further promote knowledge sharing in multilingual models. In this work, we apply the domain adversarial network to encourage the shared layers of a multilingual model to learn language-invariant features. Bidirectional Long Short-Term Memory (LSTM) recurrent neural networks (RNN) are used as building blocks. We show that shared layers learned this way contain less language identification information and lead to better performance. In an automatic speech recognition task for seven languages, the resultant acoustic model improves the word error rate (WER) of the multilingual model by 4% relative on average, and the monolingual models by 10%.</description><subject>Computer Science - Computation and Language</subject><subject>Computer Science - Learning</subject><subject>Computer Science - Sound</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2019</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj7uOwjAURN1QrIAP2Ir8QLLXOL6OyyhiYSUQBfTRxQ9kKSTIIQj-nsdSjWZGGs1h7JtDlhdSwg_FW7hmXANmoECLL6ZKe3WxpxioSfaRQhvaY-K7mGyG5hKapxueTWm6ob8Ek2w6617hhI08Nb2bfnTMdr-LfbVK19vlX1WuU0IlUq_NgRRJ0i7PORijFCCfFwYL5aW3UiuBmgqJBrwnx4W03uIBjOaAKMZs9r_6Pl6fYzhRvNcvgPoNIB5Jc0CW</recordid><startdate>20190617</startdate><enddate>20190617</enddate><creator>Hu, Ke</creator><creator>Sak, Hasim</creator><creator>Liao, Hank</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20190617</creationdate><title>Adversarial Training for Multilingual Acoustic Modeling</title><author>Hu, Ke ; Sak, Hasim ; Liao, Hank</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a673-f9cba7a5a9e4410cc7706128c687f5fd597369a856c0ffae135dfd6b0c910663</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2019</creationdate><topic>Computer Science - Computation and Language</topic><topic>Computer Science - Learning</topic><topic>Computer Science - Sound</topic><toplevel>online_resources</toplevel><creatorcontrib>Hu, Ke</creatorcontrib><creatorcontrib>Sak, Hasim</creatorcontrib><creatorcontrib>Liao, Hank</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Hu, Ke</au><au>Sak, Hasim</au><au>Liao, Hank</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Adversarial Training for Multilingual Acoustic Modeling</atitle><date>2019-06-17</date><risdate>2019</risdate><abstract>Multilingual training has been shown to improve acoustic modeling performance by sharing and transferring knowledge in modeling different languages. Knowledge sharing is usually achieved by using common lower-level layers for different languages in a deep neural network. Recently, the domain adversarial network was proposed to reduce domain mismatch of training data and learn domain-invariant features. It is thus worth exploring whether adversarial training can further promote knowledge sharing in multilingual models. In this work, we apply the domain adversarial network to encourage the shared layers of a multilingual model to learn language-invariant features. Bidirectional Long Short-Term Memory (LSTM) recurrent neural networks (RNN) are used as building blocks. We show that shared layers learned this way contain less language identification information and lead to better performance. In an automatic speech recognition task for seven languages, the resultant acoustic model improves the word error rate (WER) of the multilingual model by 4% relative on average, and the monolingual models by 10%.</abstract><doi>10.48550/arxiv.1906.07093</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.1906.07093
ispartof
issn
language eng
recordid cdi_arxiv_primary_1906_07093
source arXiv.org
subjects Computer Science - Computation and Language
Computer Science - Learning
Computer Science - Sound
title Adversarial Training for Multilingual Acoustic Modeling
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-08T19%3A40%3A44IST&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=Adversarial%20Training%20for%20Multilingual%20Acoustic%20Modeling&rft.au=Hu,%20Ke&rft.date=2019-06-17&rft_id=info:doi/10.48550/arxiv.1906.07093&rft_dat=%3Carxiv_GOX%3E1906_07093%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