@inproceedings{kleen-etal-2026-speaker,
title = "Speaker Normalization via Voice Conversion Reveals a Human-Machine Dissociation in Dialect Classification",
author = "Kleen, Caroline and
Fischbach, Lea and
Karimi, Akbar and
Flek, Lucie and
Lameli, Alfred",
editor = "Anastasopoulos, Antonis and
Markantonatou, Stella and
Ralli, Angela and
Zampieri, Marcos and
Bompolas, Stavros and
Stamou, Vivian",
booktitle = "Proceedings of the First Workshop on Dialects in {NLP} {---} A Resource Perspective",
month = may,
year = "2026",
address = "Palma de Mallorca",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.dialres-1.18/",
doi = "10.63317/3sqk7nxsikhp",
pages = "177--187",
abstract = "This study evaluates whether Retrieval-based Voice Conversion (RVC) can be used to normalize speaker-specific variability while preserving dialect-relevant acoustic cues, and what the response of human and machine systems to this manipulation reveals about the architecture of dialect recognition. In two perception experiments, speech samples from nine German dialect regions were presented either in their original form or after conversion to a single target speaker. We compared overall accuracy, confusion structures, item-level response distributions, and the interaction between listener origin and target dialect across conditions. Human classification remained stable under voice conversion. Accuracy did not differ between conditions, confusion matrices were highly correlated, and item-level divergences were minimal. The interaction between listener origin and target dialect{---}reflecting systematic regional bias{---}remained invariant. These findings indicate that RVC does not distort perceptually relevant dialectal cues and that human dialect recognition is robust to speaker normalization. In contrast, we evaluated a deep learning model under matched conditions: model accuracy improved significantly under RVC, while human performance remained unchanged. This dissociation reframes RVC as an experimental probe for investigating the divergence between human and machine speech processing, suggesting that this divergence is rooted in fundamentally different representational architectures."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kleen-etal-2026-speaker">
<titleInfo>
<title>Speaker Normalization via Voice Conversion Reveals a Human-Machine Dissociation in Dialect Classification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Caroline</namePart>
<namePart type="family">Kleen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lea</namePart>
<namePart type="family">Fischbach</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Akbar</namePart>
<namePart type="family">Karimi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lucie</namePart>
<namePart type="family">Flek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alfred</namePart>
<namePart type="family">Lameli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the First Workshop on Dialects in NLP — A Resource Perspective</title>
</titleInfo>
<name type="personal">
<namePart type="given">Antonis</namePart>
<namePart type="family">Anastasopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Stella</namePart>
<namePart type="family">Markantonatou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Angela</namePart>
<namePart type="family">Ralli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Marcos</namePart>
<namePart type="family">Zampieri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Stavros</namePart>
<namePart type="family">Bompolas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivian</namePart>
<namePart type="family">Stamou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Palma de Mallorca</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>This study evaluates whether Retrieval-based Voice Conversion (RVC) can be used to normalize speaker-specific variability while preserving dialect-relevant acoustic cues, and what the response of human and machine systems to this manipulation reveals about the architecture of dialect recognition. In two perception experiments, speech samples from nine German dialect regions were presented either in their original form or after conversion to a single target speaker. We compared overall accuracy, confusion structures, item-level response distributions, and the interaction between listener origin and target dialect across conditions. Human classification remained stable under voice conversion. Accuracy did not differ between conditions, confusion matrices were highly correlated, and item-level divergences were minimal. The interaction between listener origin and target dialect—reflecting systematic regional bias—remained invariant. These findings indicate that RVC does not distort perceptually relevant dialectal cues and that human dialect recognition is robust to speaker normalization. In contrast, we evaluated a deep learning model under matched conditions: model accuracy improved significantly under RVC, while human performance remained unchanged. This dissociation reframes RVC as an experimental probe for investigating the divergence between human and machine speech processing, suggesting that this divergence is rooted in fundamentally different representational architectures.</abstract>
<identifier type="citekey">kleen-etal-2026-speaker</identifier>
<identifier type="doi">10.63317/3sqk7nxsikhp</identifier>
<location>
<url>https://aclanthology.org/2026.dialres-1.18/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>177</start>
<end>187</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Speaker Normalization via Voice Conversion Reveals a Human-Machine Dissociation in Dialect Classification
%A Kleen, Caroline
%A Fischbach, Lea
%A Karimi, Akbar
%A Flek, Lucie
%A Lameli, Alfred
%Y Anastasopoulos, Antonis
%Y Markantonatou, Stella
%Y Ralli, Angela
%Y Zampieri, Marcos
%Y Bompolas, Stavros
%Y Stamou, Vivian
%S Proceedings of the First Workshop on Dialects in NLP — A Resource Perspective
%D 2026
%8 May
%I Association for Computational Linguistics
%C Palma de Mallorca
%F kleen-etal-2026-speaker
%X This study evaluates whether Retrieval-based Voice Conversion (RVC) can be used to normalize speaker-specific variability while preserving dialect-relevant acoustic cues, and what the response of human and machine systems to this manipulation reveals about the architecture of dialect recognition. In two perception experiments, speech samples from nine German dialect regions were presented either in their original form or after conversion to a single target speaker. We compared overall accuracy, confusion structures, item-level response distributions, and the interaction between listener origin and target dialect across conditions. Human classification remained stable under voice conversion. Accuracy did not differ between conditions, confusion matrices were highly correlated, and item-level divergences were minimal. The interaction between listener origin and target dialect—reflecting systematic regional bias—remained invariant. These findings indicate that RVC does not distort perceptually relevant dialectal cues and that human dialect recognition is robust to speaker normalization. In contrast, we evaluated a deep learning model under matched conditions: model accuracy improved significantly under RVC, while human performance remained unchanged. This dissociation reframes RVC as an experimental probe for investigating the divergence between human and machine speech processing, suggesting that this divergence is rooted in fundamentally different representational architectures.
%R 10.63317/3sqk7nxsikhp
%U https://aclanthology.org/2026.dialres-1.18/
%U https://doi.org/10.63317/3sqk7nxsikhp
%P 177-187
Markdown (Informal)
[Speaker Normalization via Voice Conversion Reveals a Human-Machine Dissociation in Dialect Classification](https://aclanthology.org/2026.dialres-1.18/) (Kleen et al., DialRes 2026)
ACL