Mixture of Phonetic Experts Based Low-Rank Adaptation of Conformer Models for Accented English Speech Recognition

查看原文 / View Original 编辑 / Edit

Computer Science > Computation and Language arXiv:2610.26888 (cs) [Submitted on 13 Oct 2026] Title: Mixture of Phonetic Experts Based Low-Rank Adaptation of Conformer Models for Accented English Speech Recognition Authors: Santosh Dahal, Anmol Guragain, Tianchi Liu, Luis Fernando D’Haro, Kiran Chandra Dahal.

计算机科学 > 计算与语言 arXiv:2610.26888 (cs) [提交于 2026 年 10 月 13 日] 标题：基于语音专家混合的 Conformer 模型低秩自适应技术用于带口音英语语音识别作者：Santosh Dahal, Anmol Guragain, Tianchi Liu, Luis Fernando D’Haro, Kiran Chandra Dahal。

Abstract: We present Mixture of Phonetic Experts based Low-Rank Adaptation (MoPE-LoRA), a parameter-efficient adaptation framework for accented English speech recognition. Recent mixture-of-experts approaches typically allocate experts per accent, implicitly treating accent identity as the primary axis of variation. In contrast, we organize adaptation along phonetic categories, motivated by the observation that accent differences are often expressed as systematic distortions of phoneme realization.

摘要：我们提出了基于语音专家混合的低秩自适应（MoPE-LoRA），这是一种用于带口音英语语音识别的参数高效自适应框架。近期的专家混合方法通常按口音分配专家，隐含地将口音身份视为变化的主要轴线。相比之下，我们根据语音类别组织自适应，其动机在于观察到口音差异通常表现为音素实现的系统性失真。

MoPE-LoRA introduces a fixed set of six low-rank experts aligned with manner-of-articulation classes and performs frame-level routing through a hybrid mechanism that combines phoneme supervision with learned acoustic gating. Because the expert set is shared across accents, the model does not require accent-specific modules and naturally extends to previously unseen accents.

MoPE-LoRA 引入了一组固定的六个低秩专家，这些专家与发音方式类别对齐，并通过结合音素监督和学习声学门控的混合机制执行帧级路由。由于专家集在不同口音间共享，该模型不需要特定于口音的模块，并能自然地扩展到先前未见的口音。

Experiments on L2-ARCTIC show that MoPE-LoRA consistently improves over single LoRA and full fine-tuning. In zero-shot evaluation where one accent is held out during training, the proposed method yields a 12.3% relative improvement over single LoRA. These results indicate that phonetic-category-based decomposition offers a scalable and transferable alternative to accent-level adaptation.

在 L2-ARCTIC 上的实验表明，MoPE-LoRA 在性能上始终优于单一 LoRA 和全参数微调。在训练期间排除一种口音的零样本评估中，该方法相对于单一 LoRA 取得了 12.3% 的相对提升。这些结果表明，基于语音类别的分解为口音级自适应提供了一种可扩展且可迁移的替代方案。