<?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-04696316</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="2026-05-22T06:28:26+02:00"/>
      </publicationStmt>
      <sourceDesc>
        <p part="N">HAL API Platform</p>
      </sourceDesc>
    </fileDesc>
  </teiHeader>
  <text>
    <body>
      <listBibl>
        <biblFull>
          <titleStmt>
            <title xml:lang="en">On the calibration of powerset speaker diarization models</title>
            <author role="aut">
              <persName>
                <forename type="first">Alexis</forename>
                <surname>Plaquet</surname>
              </persName>
              <email type="md5">3ff58ea99df9d04eda85ce2733b0f421</email>
              <email type="domain">irit.fr</email>
              <idno type="idhal" notation="string">alexis-plaquet</idno>
              <idno type="idhal" notation="numeric">1291841</idno>
              <idno type="halauthorid" notation="string">2916576-1291841</idno>
              <idno type="ORCID">https://orcid.org/0009-0004-5123-6731</idno>
              <idno type="GOOGLE SCHOLAR">https://scholar.google.fr/citations?user=7gJ465gAAAAJ</idno>
              <idno type="IDREF">https://www.idref.fr/293475245</idno>
              <affiliation ref="#struct-393117"/>
              <affiliation ref="#struct-217752"/>
            </author>
            <author role="aut">
              <persName>
                <forename type="first">Hervé</forename>
                <surname>Bredin</surname>
              </persName>
              <email type="md5">93e07d85c72c4ca943d3d32c539d08e2</email>
              <email type="domain">irit.fr</email>
              <idno type="idhal" notation="string">hbredin</idno>
              <idno type="idhal" notation="numeric">15856</idno>
              <idno type="halauthorid" notation="string">10397-15856</idno>
              <idno type="ORCID">https://orcid.org/0000-0002-3739-925X</idno>
              <idno type="GOOGLE SCHOLAR">https://scholar.google.fr/citations?user=obEwWjZOI0cC</idno>
              <idno type="IDREF">https://www.idref.fr/121165779</idno>
              <affiliation ref="#struct-393117"/>
              <affiliation ref="#struct-441569"/>
            </author>
            <editor role="depositor">
              <persName>
                <forename>Alexis</forename>
                <surname>Plaquet</surname>
              </persName>
              <email type="md5">57710a2508704ae67e4fb6059d751e20</email>
              <email type="domain">plaquet.me</email>
            </editor>
          </titleStmt>
          <editionStmt>
            <edition n="v1" type="current">
              <date type="whenSubmitted">2024-09-24 02:40:03</date>
              <date type="whenModified">2025-10-22 18:04:09</date>
              <date type="whenReleased">2024-09-24 10:57:05</date>
              <date type="whenProduced">2024-09-01</date>
              <date type="whenEndEmbargoed">2024-09-24</date>
              <ref type="file" target="https://hal.science/hal-04696316v1/document">
                <date notBefore="2024-09-24"/>
              </ref>
              <ref type="file" subtype="author" n="1" target="https://hal.science/hal-04696316v1/file/template.pdf" id="file-4696318-4094148">
                <date notBefore="2024-09-24"/>
              </ref>
            </edition>
            <respStmt>
              <resp>contributor</resp>
              <name key="1424536">
                <persName>
                  <forename>Alexis</forename>
                  <surname>Plaquet</surname>
                </persName>
                <email type="md5">57710a2508704ae67e4fb6059d751e20</email>
                <email type="domain">plaquet.me</email>
              </name>
            </respStmt>
          </editionStmt>
          <publicationStmt>
            <distributor>CCSD</distributor>
            <idno type="halId">hal-04696316</idno>
            <idno type="halUri">https://hal.science/hal-04696316</idno>
            <idno type="halBibtex">plaquet:hal-04696316</idno>
            <idno type="halRefHtml">&lt;i&gt;25th Interspeech Conference (INTERSPEECH 2024)&lt;/i&gt;, ISCA: International Speech Communication Association, Sep 2024, Kos, Greece. pp.3764--3768, &lt;a target="_blank" href="https://dx.doi.org/10.21437/Interspeech.2024-1060"&gt;&amp;#x27E8;10.21437/Interspeech.2024-1060&amp;#x27E9;&lt;/a&gt;</idno>
            <idno type="halRef">25th Interspeech Conference (INTERSPEECH 2024), ISCA: International Speech Communication Association, Sep 2024, Kos, Greece. pp.3764--3768, &amp;#x27E8;10.21437/Interspeech.2024-1060&amp;#x27E9;</idno>
            <availability status="restricted">
              <licence target="https://creativecommons.org/licenses/by/4.0/">CC BY 4.0 - Attribution<ref corresp="#file-4696318-4094148"/></licence>
            </availability>
          </publicationStmt>
          <seriesStmt>
            <idno type="stamp" n="UNIV-TLSE2">Université Toulouse 2</idno>
            <idno type="stamp" n="UNIV-TLSE3">Université de Toulouse</idno>
            <idno type="stamp" n="CNRS">CNRS - Centre national de la recherche scientifique</idno>
            <idno type="stamp" n="UT1-CAPITOLE">Université Toulouse 1 Capitole</idno>
            <idno type="stamp" n="GENCI">Publications ayant eu recours aux supercalculateurs du GENCI</idno>
            <idno type="stamp" n="IRIT">IRIT - Institut de Recherche en Informatique de Toulouse</idno>
            <idno type="stamp" n="IRIT-SAMOVA" corresp="IRIT">Structuration, Analyse et Modélisation de documents Vidéo et Audio</idno>
            <idno type="stamp" n="IRIT-SI" corresp="IRIT">Signaux et Images</idno>
            <idno type="stamp" n="IRIT-CNRS">IRIT - Centre National de la Recherche Scientifique</idno>
            <idno type="stamp" n="IRIT-UT3">IRIT - Université Toulouse III Paul Sabatier</idno>
            <idno type="stamp" n="TOULOUSE-INP">Toulouse - INP</idno>
            <idno type="stamp" n="UNIV-UT3">Université Toulouse 3</idno>
            <idno type="stamp" n="UT3-INP">Université de Toulouse / Toulouse INP</idno>
            <idno type="stamp" n="UT3-TOULOUSEINP">Université de Toulouse / Toulouse INP</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="1">Yes</note>
          </notesStmt>
          <sourceDesc>
            <biblStruct>
              <analytic>
                <title xml:lang="en">On the calibration of powerset speaker diarization models</title>
                <author role="aut">
                  <persName>
                    <forename type="first">Alexis</forename>
                    <surname>Plaquet</surname>
                  </persName>
                  <email type="md5">3ff58ea99df9d04eda85ce2733b0f421</email>
                  <email type="domain">irit.fr</email>
                  <idno type="idhal" notation="string">alexis-plaquet</idno>
                  <idno type="idhal" notation="numeric">1291841</idno>
                  <idno type="halauthorid" notation="string">2916576-1291841</idno>
                  <idno type="ORCID">https://orcid.org/0009-0004-5123-6731</idno>
                  <idno type="GOOGLE SCHOLAR">https://scholar.google.fr/citations?user=7gJ465gAAAAJ</idno>
                  <idno type="IDREF">https://www.idref.fr/293475245</idno>
                  <affiliation ref="#struct-393117"/>
                  <affiliation ref="#struct-217752"/>
                </author>
                <author role="aut">
                  <persName>
                    <forename type="first">Hervé</forename>
                    <surname>Bredin</surname>
                  </persName>
                  <email type="md5">93e07d85c72c4ca943d3d32c539d08e2</email>
                  <email type="domain">irit.fr</email>
                  <idno type="idhal" notation="string">hbredin</idno>
                  <idno type="idhal" notation="numeric">15856</idno>
                  <idno type="halauthorid" notation="string">10397-15856</idno>
                  <idno type="ORCID">https://orcid.org/0000-0002-3739-925X</idno>
                  <idno type="GOOGLE SCHOLAR">https://scholar.google.fr/citations?user=obEwWjZOI0cC</idno>
                  <idno type="IDREF">https://www.idref.fr/121165779</idno>
                  <affiliation ref="#struct-393117"/>
                  <affiliation ref="#struct-441569"/>
                </author>
              </analytic>
              <monogr>
                <idno type="isbn">ISSN: 2958-1796</idno>
                <title level="m">INTERSPEECH 2024</title>
                <meeting>
                  <title>25th Interspeech Conference (INTERSPEECH 2024)</title>
                  <date type="start">2024-09-01</date>
                  <date type="end">2024-09-05</date>
                  <settlement>Kos</settlement>
                  <country key="GR">Greece</country>
                </meeting>
                <respStmt>
                  <resp>conferenceOrganizer</resp>
                  <name>ISCA: International Speech Communication Association</name>
                </respStmt>
                <imprint>
                  <publisher>ISCA:  International Speech Communication Association</publisher>
                  <biblScope unit="pp">3764--3768</biblScope>
                  <date type="datePub">2024</date>
                </imprint>
              </monogr>
              <idno type="doi">10.21437/Interspeech.2024-1060</idno>
              <ref target="https://www.isca-archive.org/interspeech_2024/index.html" type="seeAlso"/>
              <relatedItem target="https://doi.org/10.5281/zenodo.12747242" type="Cites" subtype="http://purl.org/coar/resource_type/c_2f33"/>
            </biblStruct>
          </sourceDesc>
          <profileDesc>
            <langUsage>
              <language ident="en">English</language>
            </langUsage>
            <textClass>
              <keywords scheme="author">
                <term xml:lang="en">Speaker diarization</term>
                <term xml:lang="en">Calibration</term>
                <term xml:lang="en">Powerset classification</term>
                <term xml:lang="en">Confidence Estimation</term>
              </keywords>
              <classCode scheme="halDomain" n="info.info-ai">Computer Science [cs]/Artificial Intelligence [cs.AI]</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>End-to-end neural diarization models have usually relied on a multilabel-classification formulation of the speaker diarization problem. Recently, we proposed a powerset multiclass formulation that has beaten the state-of-the-art on multiple datasets. In this paper, we propose to study the calibration of a powerset speaker diarization model, and explore some of its uses. We study the calibration in-domain, as well as out-of-domain, and explore the data in low-confidence regions. The reliability of model confidence is then tested in practice: we use the confidence of the pretrained model to selectively create training and validation subsets out of unannotated data, and compare this to random selection. We find that top-label confidence can be used to reliably predict high-error regions. Moreover, training on low-confidence regions provides a better calibrated model, and validating on low-confidence regions can be more annotation-efficient than random regions.</p>
            </abstract>
          </profileDesc>
        </biblFull>
      </listBibl>
    </body>
    <back>
      <listOrg type="structures">
        <org type="researchteam" xml:id="struct-393117" status="OLD">
          <orgName>Équipe Structuration, Analyse et MOdélisation de documents Vidéo et Audio</orgName>
          <orgName type="acronym">IRIT-SAMoVA</orgName>
          <date type="end">2025-01-01</date>
          <desc>
            <address>
              <addrLine>IRIT118 Route de Narbonne31062 Toulouse Cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">https://www.irit.fr/SAMOVA/site/</ref>
          </desc>
          <listRelation>
            <relation active="#struct-34499" type="direct"/>
            <relation active="#struct-81148" type="indirect"/>
            <relation name="443875" active="#struct-443875" type="indirect"/>
            <relation active="#struct-116256" type="indirect"/>
            <relation active="#struct-443875" type="indirect"/>
            <relation active="#struct-217752" type="indirect"/>
            <relation name="UMR5505 / URA1399" active="#struct-441569" type="indirect"/>
            <relation active="#struct-448187" type="indirect"/>
            <relation active="#struct-1149079" type="indirect"/>
            <relation name="FED4171" active="#struct-217752" type="indirect"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-217752" status="OLD">
          <idno type="IdRef">026404672</idno>
          <idno type="ISNI">0000000121617331</idno>
          <idno type="ROR">https://ror.org/02v6kpv12</idno>
          <idno type="Wikidata">Q1273188</idno>
          <orgName>Université Toulouse III - Paul Sabatier</orgName>
          <orgName type="acronym">UT3</orgName>
          <date type="end">2025-01-01</date>
          <desc>
            <address>
              <addrLine>118 route de Narbonne - 31062 Toulouse</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.univ-tlse3.fr/</ref>
          </desc>
          <listRelation>
            <relation active="#struct-443875" type="direct"/>
          </listRelation>
        </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>
        <org type="laboratory" xml:id="struct-34499" status="OLD">
          <idno type="IdRef">030947448</idno>
          <idno type="ISNI">0000000121927225</idno>
          <idno type="RNSR">199511949P</idno>
          <idno type="ROR">https://ror.org/01rx4qw44</idno>
          <orgName>Institut de recherche en informatique de Toulouse</orgName>
          <orgName type="acronym">IRIT</orgName>
          <date type="start">1990-01-01</date>
          <date type="end">2025-01-01</date>
          <desc>
            <address>
              <addrLine>118 Route de Narbonne, F-31062 Toulouse Cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.irit.fr/</ref>
          </desc>
          <listRelation>
            <relation active="#struct-81148" type="direct"/>
            <relation name="443875" active="#struct-443875" type="indirect"/>
            <relation active="#struct-116256" type="direct"/>
            <relation active="#struct-443875" type="indirect"/>
            <relation active="#struct-217752" type="direct"/>
            <relation name="UMR5505 / URA1399" active="#struct-441569" type="direct"/>
            <relation active="#struct-448187" type="direct"/>
            <relation active="#struct-1149079" type="direct"/>
            <relation name="FED4171" active="#struct-217752" type="direct"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-81148" status="VALID">
          <idno type="IdRef">026404354</idno>
          <idno type="ISNI">0000000121902837</idno>
          <idno type="ROR">https://ror.org/0443n9e75</idno>
          <orgName>Université Toulouse Capitole</orgName>
          <orgName type="acronym">UT Capitole</orgName>
          <date type="start">1970-01-01</date>
          <desc>
            <address>
              <addrLine>2 rue du Doyen-Gabriel-Marty - 31042 Toulouse Cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.ut-capitole.fr/</ref>
          </desc>
          <listRelation>
            <relation name="443875" active="#struct-443875" type="direct"/>
          </listRelation>
        </org>
        <org type="regroupinstitution" xml:id="struct-443875" status="VALID">
          <idno type="ROR">https://ror.org/017tgbk05</idno>
          <orgName>Communauté d'universités et établissements de Toulouse</orgName>
          <orgName type="acronym">Comue de Toulouse</orgName>
          <desc>
            <address>
              <addrLine>41 Allée Jules Guesde, 31000 Toulouse</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">https://www.univ-toulouse.fr/</ref>
          </desc>
        </org>
        <org type="institution" xml:id="struct-116256" status="VALID">
          <idno type="IdRef">026403994</idno>
          <idno type="ROR">https://ror.org/04ezk3x31</idno>
          <orgName>Université Toulouse - Jean Jaurès</orgName>
          <orgName type="acronym">UT2J</orgName>
          <date type="start">1971-01-01</date>
          <desc>
            <address>
              <addrLine>5 allées Antonio Machado - 31058 Toulouse Cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">http://www.univ-tlse2.fr</ref>
          </desc>
          <listRelation>
            <relation active="#struct-443875" type="direct"/>
          </listRelation>
        </org>
        <org type="institution" xml:id="struct-448187" status="VALID">
          <idno type="IdRef">026388820</idno>
          <idno type="ISNI">0000000121766169</idno>
          <idno type="ROR">https://ror.org/033p9g875</idno>
          <orgName>Institut National Polytechnique (Toulouse)</orgName>
          <orgName type="acronym">Toulouse INP</orgName>
          <desc>
            <address>
              <country key="FR"/>
            </address>
            <ref type="url">https://www.inp-toulouse.fr/</ref>
          </desc>
          <listRelation>
            <relation active="#struct-443875" type="direct"/>
          </listRelation>
        </org>
        <org type="regrouplaboratory" xml:id="struct-1149079" status="OLD">
          <orgName>Toulouse Mind &amp; Brain Institut</orgName>
          <orgName type="acronym">TMBI</orgName>
          <date type="start">2021-01-01</date>
          <date type="end">2025-01-01</date>
          <desc>
            <address>
              <addrLine>CHU Purpan, Pavillon Baudot, 31059 Toulouse cedex 9</addrLine>
              <country key="FR"/>
            </address>
            <ref type="url">https://www.tmbi.fr/</ref>
          </desc>
          <listRelation>
            <relation active="#struct-116256" type="direct"/>
            <relation active="#struct-443875" type="indirect"/>
            <relation name="FED4171" active="#struct-217752" type="direct"/>
          </listRelation>
        </org>
      </listOrg>
    </back>
  </text>
</TEI>