<?xml version="1.0" encoding="utf-8"?>
<TEI xmlns="http://www.tei-c.org/ns/1.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:hal="http://hal.archives-ouvertes.fr/" xmlns:gml="http://www.opengis.net/gml/3.3/" xmlns:gmlce="http://www.opengis.net/gml/3.3/ce" version="1.1" xsi:schemaLocation="http://www.tei-c.org/ns/1.0 http://api.archives-ouvertes.fr/documents/aofr-sword.xsd">
  <teiHeader>
    <fileDesc>
      <titleStmt>
        <title>HAL TEI export of hal-02912029</title>
      </titleStmt>
      <publicationStmt>
        <distributor>CCSD</distributor>
        <availability status="restricted">
          <licence target="https://creativecommons.org/publicdomain/zero/1.0/">CC0 1.0 - Universal</licence>
        </availability>
        <date when="2026-05-19T16:14:48+02:00"/>
      </publicationStmt>
      <sourceDesc>
        <p part="N">HAL API Platform</p>
      </sourceDesc>
    </fileDesc>
  </teiHeader>
  <text>
    <body>
      <listBibl>
        <biblFull>
          <titleStmt>
            <title xml:lang="en">A Convolutional Deep Markov Model for Unsupervised Speech Representation Learning</title>
            <author role="aut">
              <persName>
                <forename type="first">Sameer</forename>
                <surname>Khurana</surname>
              </persName>
              <email type="md5">d45580d0a858e6194dc169169b0a2d57</email>
              <email type="domain">mit.edu</email>
              <idno type="idhal" notation="numeric">1075440</idno>
              <idno type="halauthorid" notation="string">2008278-1075440</idno>
              <affiliation ref="#struct-1050757"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Antoine</forename>
                <surname>Laurent</surname>
              </persName>
              <email type="md5">58703a365f961c551581ea38233d8ddc</email>
              <email type="domain">univ-lemans.fr</email>
              <idno type="idhal" notation="string">antoine-laurent</idno>
              <idno type="idhal" notation="numeric">13586</idno>
              <idno type="halauthorid" notation="string">33154-13586</idno>
              <idno type="ORCID">https://orcid.org/0000-0002-2653-1008</idno>
              <idno type="IDREF">https://www.idref.fr/147099072</idno>
              <affiliation ref="#struct-117974"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Wei-Ning</forename>
                <surname>Hsu</surname>
              </persName>
              <idno type="halauthorid">2008280-0</idno>
              <affiliation ref="#struct-1050757"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Jan</forename>
                <surname>Chorowski</surname>
              </persName>
              <idno type="halauthorid">1739582-0</idno>
              <affiliation ref="#struct-305685"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Adrian</forename>
                <surname>Łańcucki</surname>
              </persName>
              <idno type="halauthorid">1739584-0</idno>
              <affiliation ref="#struct-305685"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Ricard</forename>
                <surname>Marxer</surname>
              </persName>
              <email type="md5">5acbab6a8b091e3c949098a46a2ba92b</email>
              <email type="domain">univ-tln.fr</email>
              <idno type="idhal" notation="string">ricard-marxer</idno>
              <idno type="idhal" notation="numeric">19391</idno>
              <idno type="halauthorid" notation="string">27165-19391</idno>
              <idno type="ORCID">https://orcid.org/0000-0001-5099-5059</idno>
              <idno type="IDREF">https://www.idref.fr/240437713</idno>
              <affiliation ref="#struct-530702"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">James</forename>
                <surname>Glass</surname>
              </persName>
              <idno type="halauthorid">243161-0</idno>
              <affiliation ref="#struct-1050757"/>
            </author>
            <editor role="depositor">
              <persName>
                <forename>Antoine</forename>
                <surname>LAURENT</surname>
              </persName>
              <email type="md5">58703a365f961c551581ea38233d8ddc</email>
              <email type="domain">univ-lemans.fr</email>
            </editor>
          </titleStmt>
          <editionStmt>
            <edition n="v1" type="current">
              <date type="whenSubmitted">2020-08-05 09:49:18</date>
              <date type="whenModified">2026-05-19 14:10:09</date>
              <date type="whenReleased">2020-08-05 14:10:55</date>
              <date type="whenProduced">2020-10-25</date>
              <date type="whenEndEmbargoed">2020-08-05</date>
              <ref type="file" target="https://hal.science/hal-02912029v1/document">
                <date notBefore="2020-08-05"/>
              </ref>
              <ref type="file" subtype="author" n="1" target="https://hal.science/hal-02912029v1/file/convDMM_arxiv.pdf" id="file-2912029-2564012">
                <date notBefore="2020-08-05"/>
              </ref>
            </edition>
            <respStmt>
              <resp>contributor</resp>
              <name key="492760">
                <persName>
                  <forename>Antoine</forename>
                  <surname>LAURENT</surname>
                </persName>
                <email type="md5">58703a365f961c551581ea38233d8ddc</email>
                <email type="domain">univ-lemans.fr</email>
              </name>
            </respStmt>
          </editionStmt>
          <publicationStmt>
            <distributor>CCSD</distributor>
            <idno type="halId">hal-02912029</idno>
            <idno type="halUri">https://hal.science/hal-02912029</idno>
            <idno type="halBibtex">khurana:hal-02912029</idno>
            <idno type="halRefHtml">&lt;i&gt;Interspeech 2020&lt;/i&gt;, Oct 2020, Shanghai, China</idno>
            <idno type="halRef">Interspeech 2020, Oct 2020, Shanghai, China</idno>
            <availability status="restricted">
              <licence target="https://about.hal.science/hal-authorisation-v1/">HAL Authorization<ref corresp="#file-2912029-2564012"/></licence>
            </availability>
          </publicationStmt>
          <seriesStmt>
            <idno type="stamp" n="UNIV-TLN">Université de Toulon</idno>
            <idno type="stamp" n="CNRS">CNRS - Centre national de la recherche scientifique</idno>
            <idno type="stamp" n="UNIV-AMU">Aix Marseille Université</idno>
            <idno type="stamp" n="UNIV-LEMANS">Université du Mans</idno>
            <idno type="stamp" n="LIUM" corresp="UNIV-LEMANS">Laboratoire Informatique de l'Université du Maine</idno>
            <idno type="stamp" n="LIS-LAB">Laboratoire d'Informatique et Systèmes</idno>
          </seriesStmt>
          <notesStmt>
            <note type="audience" n="2">International</note>
            <note type="invited" n="0">No</note>
            <note type="popular" n="0">No</note>
            <note type="peer" n="1">Yes</note>
            <note type="proceedings" n="0">No</note>
          </notesStmt>
          <sourceDesc>
            <biblStruct>
              <analytic>
                <title xml:lang="en">A Convolutional Deep Markov Model for Unsupervised Speech Representation Learning</title>
                <author role="aut">
                  <persName>
                    <forename type="first">Sameer</forename>
                    <surname>Khurana</surname>
                  </persName>
                  <email type="md5">d45580d0a858e6194dc169169b0a2d57</email>
                  <email type="domain">mit.edu</email>
                  <idno type="idhal" notation="numeric">1075440</idno>
                  <idno type="halauthorid" notation="string">2008278-1075440</idno>
                  <affiliation ref="#struct-1050757"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Antoine</forename>
                    <surname>Laurent</surname>
                  </persName>
                  <email type="md5">58703a365f961c551581ea38233d8ddc</email>
                  <email type="domain">univ-lemans.fr</email>
                  <idno type="idhal" notation="string">antoine-laurent</idno>
                  <idno type="idhal" notation="numeric">13586</idno>
                  <idno type="halauthorid" notation="string">33154-13586</idno>
                  <idno type="ORCID">https://orcid.org/0000-0002-2653-1008</idno>
                  <idno type="IDREF">https://www.idref.fr/147099072</idno>
                  <affiliation ref="#struct-117974"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Wei-Ning</forename>
                    <surname>Hsu</surname>
                  </persName>
                  <idno type="halauthorid">2008280-0</idno>
                  <affiliation ref="#struct-1050757"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Jan</forename>
                    <surname>Chorowski</surname>
                  </persName>
                  <idno type="halauthorid">1739582-0</idno>
                  <affiliation ref="#struct-305685"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Adrian</forename>
                    <surname>Łańcucki</surname>
                  </persName>
                  <idno type="halauthorid">1739584-0</idno>
                  <affiliation ref="#struct-305685"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Ricard</forename>
                    <surname>Marxer</surname>
                  </persName>
                  <email type="md5">5acbab6a8b091e3c949098a46a2ba92b</email>
                  <email type="domain">univ-tln.fr</email>
                  <idno type="idhal" notation="string">ricard-marxer</idno>
                  <idno type="idhal" notation="numeric">19391</idno>
                  <idno type="halauthorid" notation="string">27165-19391</idno>
                  <idno type="ORCID">https://orcid.org/0000-0001-5099-5059</idno>
                  <idno type="IDREF">https://www.idref.fr/240437713</idno>
                  <affiliation ref="#struct-530702"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">James</forename>
                    <surname>Glass</surname>
                  </persName>
                  <idno type="halauthorid">243161-0</idno>
                  <affiliation ref="#struct-1050757"/>
                </author>
              </analytic>
              <monogr>
                <meeting>
                  <title>Interspeech 2020</title>
                  <date type="start">2020-10-25</date>
                  <settlement>Shanghai</settlement>
                  <country key="CN">China</country>
                </meeting>
                <imprint/>
              </monogr>
            </biblStruct>
          </sourceDesc>
          <profileDesc>
            <langUsage>
              <language ident="en">English</language>
            </langUsage>
            <textClass>
              <keywords scheme="author">
                <term xml:lang="en">Unsupervised Speech Representation Learning</term>
                <term xml:lang="en">Structured Variational Inference</term>
                <term xml:lang="en">Neural Variational Latent Variable Model</term>
              </keywords>
              <classCode scheme="halDomain" n="info.info-ai">Computer Science [cs]/Artificial Intelligence [cs.AI]</classCode>
              <classCode scheme="halDomain" n="info.info-cl">Computer Science [cs]/Computation and Language [cs.CL]</classCode>
              <classCode scheme="halDomain" n="info.info-lg">Computer Science [cs]/Machine Learning [cs.LG]</classCode>
              <classCode scheme="halDomain" n="info.info-ne">Computer Science [cs]/Neural and Evolutionary Computing [cs.NE]</classCode>
              <classCode scheme="halTypology" n="COMM">Conference papers</classCode>
              <classCode scheme="halOldTypology" n="COMM">Conference papers</classCode>
              <classCode scheme="halTreeTypology" n="COMM">Conference papers</classCode>
            </textClass>
            <abstract xml:lang="en">
              <p>Probabilistic Latent Variable Models (LVMs) provide an alternative to self-supervised learning approaches for linguistic representation learning from speech. LVMs admit an intuitive probabilistic interpretation where the latent structure shapes the information extracted from the signal. Even though LVMs have recently seen a renewed interest due to the introduction of Vari-ational Autoencoders (VAEs), their use for speech representation learning remains largely unexplored. In this work, we propose Convolutional Deep Markov Model (ConvDMM), a Gaus-sian state-space model with non-linear emission and transition functions modelled by deep neural networks. This unsupervised model is trained using black box variational inference. A deep convolutional neural network is used as an inference network for structured variational approximation. When trained on a large scale speech dataset (LibriSpeech), ConvDMM produces features that significantly outperform multiple self-supervised feature extracting methods on linear phone classification and recognition on the Wall Street Journal dataset. Furthermore, we found that ConvDMM complements self-supervised methods like Wav2Vec and PASE, improving on the results achieved with any of the methods alone. Lastly, we find that ConvDMM features enable learning better phone recognizers than any other features in an extreme low-resource regime with few labelled training examples.</p>
            </abstract>
          </profileDesc>
        </biblFull>
      </listBibl>
    </body>
    <back>
      <listOrg type="structures">
        <org type="laboratory" xml:id="struct-1050757" status="VALID">
          <orgName>MIT Computer Science &amp; Artificial Intelligence Lab</orgName>
          <orgName type="acronym">MIT CSAIL</orgName>
          <desc>
            <address>
              <country key="US"/>
            </address>
            <ref type="url">https://www.csail.mit.edu</ref>
          </desc>
          <listRelation>
            <relation active="#struct-301950" type="direct"/>
          </listRelation>
        </org>
        <org type="laboratory" xml:id="struct-117974" status="VALID">
          <idno type="IdRef">035240776</idno>
          <idno type="RNSR">200615275G</idno>
          <idno type="ROR">https://ror.org/03gvgnv89</idno>
          <idno type="Wikidata">Q51779666</idno>
          <orgName>Laboratoire d'Informatique de l'Université du Mans</orgName>
          <orgName type="acronym">LIUM</orgName>
          <desc>
            <address>
              <addrLine>Avenue Laennec 72085 Le Mans cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">https://lium.univ-lemans.fr/</ref>
          </desc>
          <listRelation>
            <relation name="EA4023" active="#struct-7566" type="direct"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-305685" status="VALID">
          <idno type="IdRef">026572346</idno>
          <idno type="ISNI">0000000121853326</idno>
          <idno type="ROR">https://ror.org/00yae6e25</idno>
          <idno type="Wikidata">Q315658</idno>
          <orgName>University of Wrocław [Poland] = Uniwersytet Wrocławski [Polska] = Universität Breslau [Polen] = Université de Wrocław [Pologne]</orgName>
          <orgName type="acronym">UWr</orgName>
          <date type="start">1702-01-01</date>
          <desc>
            <address>
              <addrLine>Place Uniwersytecki 1, 50-137 Wrocław, Pologne</addrLine>
              <country key="PL"/>
            </address>
            <ref type="url">https://uwr.edu.pl</ref>
          </desc>
        </org>
        <org type="researchteam" xml:id="struct-530702" status="VALID">
          <orgName>DYNamiques de l’Information</orgName>
          <orgName type="acronym">DYNI</orgName>
          <desc>
            <address>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.lis-lab.fr/dyni</ref>
          </desc>
          <listRelation>
            <relation active="#struct-527033" type="direct"/>
            <relation name="UMR7020" active="#struct-198056" type="indirect"/>
            <relation name="UMR7020" active="#struct-303091" type="indirect"/>
            <relation name="UMR7020" active="#struct-441569" type="indirect"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-301950" status="VALID">
          <idno type="ROR">https://ror.org/042nb2s44</idno>
          <orgName>Massachusetts Institute of Technology</orgName>
          <orgName type="acronym">MIT</orgName>
          <desc>
            <address>
              <addrLine>77 Massachusetts Ave, Cambridge, MA 02139</addrLine>
              <country key="US"/>
            </address>
            <ref type="url">http://web.mit.edu</ref>
          </desc>
        </org>
        <org type="institution" xml:id="struct-7566" status="VALID">
          <idno type="IdRef">026404435</idno>
          <idno type="ROR">https://ror.org/01mtcc283</idno>
          <orgName>Le Mans Université</orgName>
          <orgName type="acronym">UM</orgName>
          <desc>
            <address>
              <addrLine>Avenue Olivier Messiaen - 72085 Le Mans cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.univ-lemans.fr/</ref>
          </desc>
        </org>
        <org type="laboratory" xml:id="struct-527033" status="VALID">
          <idno type="IdRef">234768002</idno>
          <idno type="RNSR">201822714R</idno>
          <idno type="ROR">https://ror.org/0257sgk90</idno>
          <idno type="Wikidata">Q52605252</idno>
          <orgName>Laboratoire d'Informatique et des Systèmes (LIS) (Marseille, Toulon)</orgName>
          <orgName type="acronym">LIS</orgName>
          <date type="start">2018-01-01</date>
          <desc>
            <address>
              <addrLine>Aix Marseille Université – Campus de Saint Jérôme – Bat. Polytech, 52 Av. Escadrille Normandie Niemen, 13397 Marseille Cedex 20</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.lis-lab.fr</ref>
          </desc>
          <listRelation>
            <relation name="UMR7020" active="#struct-198056" type="direct"/>
            <relation name="UMR7020" active="#struct-303091" type="direct"/>
            <relation name="UMR7020" active="#struct-441569" type="direct"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-198056" status="VALID">
          <idno type="IdRef">15863621X</idno>
          <idno type="ISNI">0000 0001 2176 4817</idno>
          <idno type="ROR">https://ror.org/035xkbk20</idno>
          <orgName>Aix Marseille Université</orgName>
          <orgName type="acronym">AMU</orgName>
          <date type="start">2012-01-01</date>
          <desc>
            <address>
              <addrLine>Aix-Marseille UniversitéJardins du Pharo58 Boulevard Charles Livon13284 Marseille cedex 7</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.univ-amu.fr/</ref>
          </desc>
        </org>
        <org type="institution" xml:id="struct-303091" status="VALID">
          <idno type="IdRef">031122337</idno>
          <idno type="ROR">https://ror.org/02m9kbe37</idno>
          <orgName>Université de Toulon</orgName>
          <orgName type="acronym">UTLN</orgName>
          <desc>
            <address>
              <addrLine>Avenue de l'Université - BP 20132 - 83957 La Garde Cedex</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.univ-tln.fr/</ref>
          </desc>
        </org>
        <org type="regroupinstitution" xml:id="struct-441569" status="VALID">
          <idno type="IdRef">02636817X</idno>
          <idno type="ISNI">0000000122597504</idno>
          <idno type="ROR">https://ror.org/02feahw73</idno>
          <orgName>Centre National de la Recherche Scientifique</orgName>
          <orgName type="acronym">CNRS</orgName>
          <date type="start">1939-10-19</date>
          <desc>
            <address>
              <country key="FR"/>
            </address>
            <ref type="url">https://www.cnrs.fr/</ref>
          </desc>
        </org>
      </listOrg>
    </back>
  </text>
</TEI>