Adapting Where It Matters: Depth-Aware Adaptation for Efficient Multilingual Speech Recognition in Low-Resource Languages

2026-02-01

AI Summary

This work addresses the challenges of low-resource multilingual speech recognition, which are primarily constrained by data scarcity and inefficient adaptation. The authors propose DAMA, a depth-aware adaptation framework that reveals and leverages a U-shaped adaptability pattern within multilingual ASR models—allocating adaptation capacity to shallow and deep layers while freezing intermediate layers to preserve shared semantic representations. By integrating SVD-based initialization with parameter-efficient fine-tuning, DAMA achieves state-of-the-art or comparable performance across 18 low-resource languages using only 20% trainable parameters. Notably, it reduces word error rates by up to 29% under extreme data scarcity, substantially improving memory efficiency, training time, and computational cost.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Recent speech foundation models excel at multilingual automatic speech recognition (ASR) for high-resource languages, but adapting them to low-resource languages remains challenging due to data scarcity and efficiency constraints. Full-model fine-tuning is computationally expensive and prone to overfitting, while parameter-efficient methods like LoRA apply adaptation uniformly across layers, overlooking internal representations thus compromising effectiveness and efficiency. We analyze multilingual ASR models and reveal a U-shaped adaptability pattern: early and late layers are language-specific and require more adaptation, while intermediate layers retain shared semantics and need less. Building on this observation, we propose DAMA, a Depth-Aware Model Adaptation framework that allocates adaptation capacity according to each layer's role. DAMA also introduces Singular Value Decomposition (SVD)-based initialization to constrain adaptation and preserve the U-shaped pattern, as well as a frozen middle-layer basis for further efficiency. Evaluated on 18 low-resource languages across two benchmark datasets, DAMA matches or surpasses state-of-the-art accuracy with 80% fewer trainable parameters, achieves a 29% error reduction under extreme data scarcity, and significantly improves memory, training time, and computational efficiency over baselines. These results highlight the benefits of structure-aware adaptation for efficient, scalable multilingual ASR.

Problem

• low-resource languages
• multilingual speech recognition
• parameter-efficient adaptation
• depth-aware adaptation
• speech foundation models

Innovation

• Depth-Aware Adaptation
• Low-Resource ASR
• Parameter-Efficient Fine-Tuning
• U-Shaped Adaptability
• Singular Value Decomposition