@inproceedings{bengono-obiang-etal-2026-dynamic,
title = "Dynamic Layer Selection for Efficient Tone Recognition in Self-Supervised Speech Models",
author = "BENGONO OBIANG, Saint Germes B. and
TSOPZE, Norbert and
MELATAGIA YONTA, Paulin",
editor = "Piperidis, Stelios and
Bel, N{\'u}ria and
van den Heuvel, Henk and
Ide, Nancy and
Krek, Simon and
Toral, Antonio",
booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
month = may,
year = "2026",
address = "Palma de Mallorca, Spain",
publisher = "ELRA Language Resource Association",
url = "https://aclanthology.org/2026.lrec-1.792/",
doi = "10.63317/3hi6ut7agp4y",
pages = "10093--10101",
abstract = "Low-resource tonal languages present significant challenges to speech processing technologies, due to limited training data and the critical role of pitch variation in expressing meaning. This paper applies established weighted layer combination methods to tone recognition in such languages, with a specific focus on Yoruba and Yemba. Building on our previous work with Wav2vec 2.0 representations and the weighted-sum methodology from Yang et al. (2024), we investigate layer specialisation in the SSA-HuBERT self-supervised speech model for tonal tasks. Our systematic analysis reveals significant performance differences between different layers, with middle layers generally outperforming both lower and upper layers for tonal recognition tasks. While typical approaches only use the output of the last layer, our experiments show that weighted layer combination outperforms the last layer by 20.4{\%} and 15.8{\%} relative improvement in tone error rate (TER) for Yoruba and Yemba, respectively. In addition to performance improvements, our approach provides dramatic computational efficiency gains, reducing the resources required by over 90{\%} compared to evaluating each layer separately. Analysis of the learned layer weights reveals language-specific patterns, with Yoruba favouring middle layers and Yemba giving more weight to early layers. These results provide valuable insights into how tonal information is encoded in self-supervised speech models, and demonstrate a practical application of established layer combination methods in low-resource language contexts."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="bengono-obiang-etal-2026-dynamic">
<titleInfo>
<title>Dynamic Layer Selection for Efficient Tone Recognition in Self-Supervised Speech Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Saint</namePart>
<namePart type="given">Germes</namePart>
<namePart type="given">B</namePart>
<namePart type="family">BENGONO OBIANG</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Norbert</namePart>
<namePart type="family">TSOPZE</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paulin</namePart>
<namePart type="family">MELATAGIA YONTA</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 Fifteenth Language Resources and Evaluation Conference</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stelios</namePart>
<namePart type="family">Piperidis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Núria</namePart>
<namePart type="family">Bel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Henk</namePart>
<namePart type="family">van den Heuvel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nancy</namePart>
<namePart type="family">Ide</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Krek</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>ELRA Language Resource Association</publisher>
<place>
<placeTerm type="text">Palma de Mallorca, Spain</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Low-resource tonal languages present significant challenges to speech processing technologies, due to limited training data and the critical role of pitch variation in expressing meaning. This paper applies established weighted layer combination methods to tone recognition in such languages, with a specific focus on Yoruba and Yemba. Building on our previous work with Wav2vec 2.0 representations and the weighted-sum methodology from Yang et al. (2024), we investigate layer specialisation in the SSA-HuBERT self-supervised speech model for tonal tasks. Our systematic analysis reveals significant performance differences between different layers, with middle layers generally outperforming both lower and upper layers for tonal recognition tasks. While typical approaches only use the output of the last layer, our experiments show that weighted layer combination outperforms the last layer by 20.4% and 15.8% relative improvement in tone error rate (TER) for Yoruba and Yemba, respectively. In addition to performance improvements, our approach provides dramatic computational efficiency gains, reducing the resources required by over 90% compared to evaluating each layer separately. Analysis of the learned layer weights reveals language-specific patterns, with Yoruba favouring middle layers and Yemba giving more weight to early layers. These results provide valuable insights into how tonal information is encoded in self-supervised speech models, and demonstrate a practical application of established layer combination methods in low-resource language contexts.</abstract>
<identifier type="citekey">bengono-obiang-etal-2026-dynamic</identifier>
<identifier type="doi">10.63317/3hi6ut7agp4y</identifier>
<location>
<url>https://aclanthology.org/2026.lrec-1.792/</url>
</location>
<part>
<date>2026-05</date>
<extent unit="page">
<start>10093</start>
<end>10101</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Dynamic Layer Selection for Efficient Tone Recognition in Self-Supervised Speech Models
%A BENGONO OBIANG, Saint Germes B.
%A TSOPZE, Norbert
%A MELATAGIA YONTA, Paulin
%Y Piperidis, Stelios
%Y Bel, Núria
%Y van den Heuvel, Henk
%Y Ide, Nancy
%Y Krek, Simon
%Y Toral, Antonio
%S Proceedings of the Fifteenth Language Resources and Evaluation Conference
%D 2026
%8 May
%I ELRA Language Resource Association
%C Palma de Mallorca, Spain
%F bengono-obiang-etal-2026-dynamic
%X Low-resource tonal languages present significant challenges to speech processing technologies, due to limited training data and the critical role of pitch variation in expressing meaning. This paper applies established weighted layer combination methods to tone recognition in such languages, with a specific focus on Yoruba and Yemba. Building on our previous work with Wav2vec 2.0 representations and the weighted-sum methodology from Yang et al. (2024), we investigate layer specialisation in the SSA-HuBERT self-supervised speech model for tonal tasks. Our systematic analysis reveals significant performance differences between different layers, with middle layers generally outperforming both lower and upper layers for tonal recognition tasks. While typical approaches only use the output of the last layer, our experiments show that weighted layer combination outperforms the last layer by 20.4% and 15.8% relative improvement in tone error rate (TER) for Yoruba and Yemba, respectively. In addition to performance improvements, our approach provides dramatic computational efficiency gains, reducing the resources required by over 90% compared to evaluating each layer separately. Analysis of the learned layer weights reveals language-specific patterns, with Yoruba favouring middle layers and Yemba giving more weight to early layers. These results provide valuable insights into how tonal information is encoded in self-supervised speech models, and demonstrate a practical application of established layer combination methods in low-resource language contexts.
%R 10.63317/3hi6ut7agp4y
%U https://aclanthology.org/2026.lrec-1.792/
%U https://doi.org/10.63317/3hi6ut7agp4y
%P 10093-10101
Markdown (Informal)
[Dynamic Layer Selection for Efficient Tone Recognition in Self-Supervised Speech Models](https://aclanthology.org/2026.lrec-1.792/) (BENGONO OBIANG et al., LREC 2026)
ACL