Industry Scale Semi-Supervised Learning for Natural Language Understanding

This paper presents a production Semi-Supervised Learning (SSL) pipeline based on the student-teacher framework, which leverages millions of unlabeled examples to improve Natural Language Understanding (NLU) tasks. We investigate two questions related to the use of unlabeled data in production SSL c...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2021-03
Hauptverfasser: Chen, Luoxin, Garcia, Francisco, Kumar, Varun, Xie, He, Lu, Jianhua
Format: Artikel
Sprache:eng
Schlagworte:
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 Chen, Luoxin
Garcia, Francisco
Kumar, Varun
Xie, He
Lu, Jianhua
description This paper presents a production Semi-Supervised Learning (SSL) pipeline based on the student-teacher framework, which leverages millions of unlabeled examples to improve Natural Language Understanding (NLU) tasks. We investigate two questions related to the use of unlabeled data in production SSL context: 1) how to select samples from a huge unlabeled data pool that are beneficial for SSL training, and 2) how do the selected data affect the performance of different state-of-the-art SSL techniques. We compare four widely used SSL techniques, Pseudo-Label (PL), Knowledge Distillation (KD), Virtual Adversarial Training (VAT) and Cross-View Training (CVT) in conjunction with two data selection methods including committee-based selection and submodular optimization based selection. We further examine the benefits and drawbacks of these techniques when applied to intent classification (IC) and named entity recognition (NER) tasks, and provide guidelines specifying when each of these methods might be beneficial to improve large scale NLU systems.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2507368772</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2507368772</sourcerecordid><originalsourceid>FETCH-proquest_journals_25073687723</originalsourceid><addsrcrecordid>eNqNykELgjAYgOERBEn5HwadhbWp8x5FhXSxzvLRPkWxad-2oH-fh35Ap_fwPgsWSaV2SZFKuWKxc70QQuZaZpmK2OVsTXCePrx6wIC8wmeXVGFCencODS8RyHa25c1I_Ao-EAy8BNsGaJHfrUFyHqyZyYYtGxgcxr-u2fZ4uO1PyUTjK6DzdT8GsvOqZSa0ygutpfpPfQHe4z0E</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2507368772</pqid></control><display><type>article</type><title>Industry Scale Semi-Supervised Learning for Natural Language Understanding</title><source>Free E- Journals</source><creator>Chen, Luoxin ; Garcia, Francisco ; Kumar, Varun ; Xie, He ; Lu, Jianhua</creator><creatorcontrib>Chen, Luoxin ; Garcia, Francisco ; Kumar, Varun ; Xie, He ; Lu, Jianhua</creatorcontrib><description>This paper presents a production Semi-Supervised Learning (SSL) pipeline based on the student-teacher framework, which leverages millions of unlabeled examples to improve Natural Language Understanding (NLU) tasks. We investigate two questions related to the use of unlabeled data in production SSL context: 1) how to select samples from a huge unlabeled data pool that are beneficial for SSL training, and 2) how do the selected data affect the performance of different state-of-the-art SSL techniques. We compare four widely used SSL techniques, Pseudo-Label (PL), Knowledge Distillation (KD), Virtual Adversarial Training (VAT) and Cross-View Training (CVT) in conjunction with two data selection methods including committee-based selection and submodular optimization based selection. We further examine the benefits and drawbacks of these techniques when applied to intent classification (IC) and named entity recognition (NER) tasks, and provide guidelines specifying when each of these methods might be beneficial to improve large scale NLU systems.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Distillation ; Natural language ; Optimization ; Semi-supervised learning ; Training</subject><ispartof>arXiv.org, 2021-03</ispartof><rights>2021. This work is published under http://creativecommons.org/licenses/by/4.0/ (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>Chen, Luoxin</creatorcontrib><creatorcontrib>Garcia, Francisco</creatorcontrib><creatorcontrib>Kumar, Varun</creatorcontrib><creatorcontrib>Xie, He</creatorcontrib><creatorcontrib>Lu, Jianhua</creatorcontrib><title>Industry Scale Semi-Supervised Learning for Natural Language Understanding</title><title>arXiv.org</title><description>This paper presents a production Semi-Supervised Learning (SSL) pipeline based on the student-teacher framework, which leverages millions of unlabeled examples to improve Natural Language Understanding (NLU) tasks. We investigate two questions related to the use of unlabeled data in production SSL context: 1) how to select samples from a huge unlabeled data pool that are beneficial for SSL training, and 2) how do the selected data affect the performance of different state-of-the-art SSL techniques. We compare four widely used SSL techniques, Pseudo-Label (PL), Knowledge Distillation (KD), Virtual Adversarial Training (VAT) and Cross-View Training (CVT) in conjunction with two data selection methods including committee-based selection and submodular optimization based selection. We further examine the benefits and drawbacks of these techniques when applied to intent classification (IC) and named entity recognition (NER) tasks, and provide guidelines specifying when each of these methods might be beneficial to improve large scale NLU systems.</description><subject>Distillation</subject><subject>Natural language</subject><subject>Optimization</subject><subject>Semi-supervised learning</subject><subject>Training</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><recordid>eNqNykELgjAYgOERBEn5HwadhbWp8x5FhXSxzvLRPkWxad-2oH-fh35Ap_fwPgsWSaV2SZFKuWKxc70QQuZaZpmK2OVsTXCePrx6wIC8wmeXVGFCencODS8RyHa25c1I_Ao-EAy8BNsGaJHfrUFyHqyZyYYtGxgcxr-u2fZ4uO1PyUTjK6DzdT8GsvOqZSa0ygutpfpPfQHe4z0E</recordid><startdate>20210329</startdate><enddate>20210329</enddate><creator>Chen, Luoxin</creator><creator>Garcia, Francisco</creator><creator>Kumar, Varun</creator><creator>Xie, He</creator><creator>Lu, Jianhua</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>20210329</creationdate><title>Industry Scale Semi-Supervised Learning for Natural Language Understanding</title><author>Chen, Luoxin ; Garcia, Francisco ; Kumar, Varun ; Xie, He ; Lu, Jianhua</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_25073687723</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Distillation</topic><topic>Natural language</topic><topic>Optimization</topic><topic>Semi-supervised learning</topic><topic>Training</topic><toplevel>online_resources</toplevel><creatorcontrib>Chen, Luoxin</creatorcontrib><creatorcontrib>Garcia, Francisco</creatorcontrib><creatorcontrib>Kumar, Varun</creatorcontrib><creatorcontrib>Xie, He</creatorcontrib><creatorcontrib>Lu, Jianhua</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>Chen, Luoxin</au><au>Garcia, Francisco</au><au>Kumar, Varun</au><au>Xie, He</au><au>Lu, Jianhua</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Industry Scale Semi-Supervised Learning for Natural Language Understanding</atitle><jtitle>arXiv.org</jtitle><date>2021-03-29</date><risdate>2021</risdate><eissn>2331-8422</eissn><abstract>This paper presents a production Semi-Supervised Learning (SSL) pipeline based on the student-teacher framework, which leverages millions of unlabeled examples to improve Natural Language Understanding (NLU) tasks. We investigate two questions related to the use of unlabeled data in production SSL context: 1) how to select samples from a huge unlabeled data pool that are beneficial for SSL training, and 2) how do the selected data affect the performance of different state-of-the-art SSL techniques. We compare four widely used SSL techniques, Pseudo-Label (PL), Knowledge Distillation (KD), Virtual Adversarial Training (VAT) and Cross-View Training (CVT) in conjunction with two data selection methods including committee-based selection and submodular optimization based selection. We further examine the benefits and drawbacks of these techniques when applied to intent classification (IC) and named entity recognition (NER) tasks, and provide guidelines specifying when each of these methods might be beneficial to improve large scale NLU systems.</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, 2021-03
issn 2331-8422
language eng
recordid cdi_proquest_journals_2507368772
source Free E- Journals
subjects Distillation
Natural language
Optimization
Semi-supervised learning
Training
title Industry Scale Semi-Supervised Learning for Natural Language Understanding
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-14T05%3A57%3A11IST&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=Industry%20Scale%20Semi-Supervised%20Learning%20for%20Natural%20Language%20Understanding&rft.jtitle=arXiv.org&rft.au=Chen,%20Luoxin&rft.date=2021-03-29&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2507368772%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2507368772&rft_id=info:pmid/&rfr_iscdi=true