MauBERT: Universal Phonetic Inductive Biases for Few-Shot Acoustic Units Discovery

2025/12/22 by Angelo Ortiz Tandazo, Manel Khentout, Tandazo, Angelo Ortiz +7 · 1 citation
Computer Science · Psychology · #Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #Emotion and Mood Recognition #FOS: Computer and information sciences #FOS: Electrical engineering #Phonetics and Phonology Research #Speech Recognition and Synthesis #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2512.19612

openalex publication_date 2025/12/22 · openalex created_date 2025/12/24 · openalex updated_date 2026/07/28

Abstract

This paper introduces MauBERT, a multilingual extension of HuBERT that leverages articulatory features for robust cross-lingual phonetic representation learning. We continue HuBERT pre-training with supervision based on a phonetic-to-articulatory feature mapping in 55 languages. Our models learn from multilingual data to predict articulatory features or phones, resulting in language-independent representations that capture multilingual phonetic properties. Through comprehensive ABX discriminability testing, we show MauBERT models produce more context-invariant representations than state-of-the-art multilingual self-supervised learning models. Additionally, the models effectively adapt to unseen languages and casual speech with minimal self-supervised fine-tuning (10 hours of speech). This establishes an effective approach for instilling linguistic inductive biases in self-supervised speech models.

Citations

• SpidR: Learning Fast and Stable Linguistic Units for Spoken Language Models Without Supervision
• fastabx: A library for efficient computation of ABX discriminability
• Simulating Early Phonetic and Word Learning Without Linguistic Categories
• Moshi: a speech-text foundation model for real-time dialogue
• Towards Robust Speech Representation Learning for Thousands of Languages
• Orthogonality and isotropy of speaker and phonetic information in self-supervised speech representations
• mHuBERT-147: A Compact Multilingual HuBERT Model
• Evaluating Self-Supervised Speech Representations for Indigenous American Languages
• Pushing the Limits of Unsupervised Unit Discovery for SSL Speech Representation
• Scaling Speech Technology to 1,000+ Languages
• ML-SUPERB: Multilingual Speech Universal PERformance Benchmark
• DinoSR: Self-Distillation and Online Clustering for Self-supervised Speech Representation Learning
• Speech-to-Speech Translation For A Real-world Unwritten Language
• Evaluating context-invariance in unsupervised speech representations
• Supervision-Guided Codebooks for Masked Prediction in Speech Pre-training
• Discovering Phonetic Inventories with Crosslingual Automatic Speech Recognition
• XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale
• HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units
• SUPERB: Speech processing Universal PERformance Benchmark
• Generative Spoken Language Modeling from Raw Audio
• UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data
• Unsupervised Cross-lingual Representation Learning for Speech\n Recognition
• wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations
• UWSpeech: Speech to Speech Translation for Unwritten Languages
• Common Voice: A Massively-Multilingual Speech Corpus
• Representation Learning with Contrastive Predictive Coding
• The Zero Resource Speech Challenge 2017
• Adam: A Method for Stochastic Optimization

Cited by

• SpidR-Adapt: A Universal Speech Representation Model for Few-Shot Adaptation

Related

• Language-Universal Phonetic Representation in Multilingual Speech Pretraining for Low-Resource Speech Recognition
• Predicting within and across language phoneme recognition performance of self-supervised learning speech pre-trained models
• That Sounds Familiar: an Analysis of Phonetic Representations Transfer\n Across Languages
• Improved ASR for Under-Resourced Languages Through Multi-Task Learning with Acoustic Landmarks
• Supervised Acoustic Embeddings And Their Transferability Across Languages