儿童语音识别适配中的成人性能保持：一项实证研究

arXiv cs.CL 论文
automatic-speech-recognition child-speech asr-adaptation lora weight-merging catastrophic-forgetting multilingual

摘要

一项实证研究，探讨在阿拉伯语和英语中将成人 ASR 模型适配到儿童语音同时保持成人性能的方法，对比了微调、LoRA 和权重空间融合（LERP/TIES）等技术，发现双语适配与权重融合能够改善适配与保持之间的权衡。

arXiv:2610.08827v1 Announce Type: new Abstract: Automatic Speech Recognition (ASR) systems often underperform for children and non-native speakers, while adapting adult ASR models to child speech can cause adult-speech forgetting. We study child ASR adaptation with adult retention across Arabic and English. We compare full fine-tuning, LoRA, and post-hoc weight-space merging across encoder--decoder, encoder--CTC, and AudioLLM-based ASR systems. Experiments use Arabic native and non-native child speech, English MyST child speech, and adult benchmarks from MGB-2 and LibriSpeech test-clean. We evaluate recognition quality with WER and quantify the adaptation--retention trade-off using Retention Index, Child Adaptation Gain, and Adaptation Recovery. Results show that child adaptation is necessary, especially for non-native Arabic and English child speech, but direct adaptation often reduces adult ASR performance. Bilingual adaptation is more stable than language-specific adaptation. Weight-space merging often improves the trade-off, especially for encoder--CTC, Whisper, and AudioLLM-based ASR, with LERP favoring adult retention and TIES recovering stronger child gains. For the encoder--decoder model, direct bilingual fine-tuning remains strongest in raw WER.\footnote{Code, and models are available at https://github.com/qcri/Child-ASR-Adaptation.
查看原文
查看缓存全文

缓存时间: 2026/10/08 12:37 

# Child ASR Adaptation with Adult Retention: An Empirical Study
Source: [https://arxiv.org/abs/2610.08827](https://arxiv.org/abs/2610.08827)
[View PDF](https://arxiv.org/pdf/2610.08827)[HTML \(experimental\)](https://arxiv.org/html/2610.08827v1)

> Abstract:Automatic Speech Recognition \(ASR\) systems often underperform for children and non\-native speakers, while adapting adult ASR models to child speech can cause adult\-speech forgetting\. We study child ASR adaptation with adult retention across Arabic and English\. We compare full fine\-tuning, LoRA, and post\-hoc weight\-space merging across encoder\-\-decoder, encoder\-\-CTC, and AudioLLM\-based ASR systems\. Experiments use Arabic native and non\-native child speech, English MyST child speech, and adult benchmarks from MGB\-2 and LibriSpeech test\-clean\. We evaluate recognition quality with WER and quantify the adaptation\-\-retention trade\-off using Retention Index, Child Adaptation Gain, and Adaptation Recovery\. Results show that child adaptation is necessary, especially for non\-native Arabic and English child speech, but direct adaptation often reduces adult ASR performance\. Bilingual adaptation is more stable than language\-specific adaptation\. Weight\-space merging often improves the trade\-off, especially for encoder\-\-CTC, Whisper, and AudioLLM\-based ASR, with LERP favoring adult retention and TIES recovering stronger child gains\. For the encoder\-\-decoder model, direct bilingual fine\-tuning remains strongest in raw WER\.\\footnote\{Code, and models are available at[this https URL](https://github.com/qcri/Child-ASR-Adaptation)\.

## Submission history

From: Houssam Eddine\-Othman Lachemat \[[view email](https://arxiv.org/show-email/1c1fc847/2610.08827)\] **\[v1\]**Fri, 25 Sep 2026 18:10:54 UTC \(651 KB\)

相似文章

使基础ASR模型适应构音障碍语音：一项案例研究

arXiv cs.CL

本文介绍了一个针对构音障碍说话者的个性化ASR系统，通过对Whisper基础模型进行微调，仅使用22.5小时的适应数据加上8.8小时的用户纠正，实现了9.7%的词错误率。结果表明，个性化微调可以显著提升基础ASR模型对构音障碍语音的效果，并且通过部署的移动应用实现了真实世界的数据收集。

转录儿童语音：ASR性能与获取可靠的正字法转写

arXiv cs.CL

这篇论文评估了九种ASR模型（Whisper、Parakeet、Wav2Vec2）在荷兰语儿童语音数据集JASMIN和DART上的表现，发现微调后的Whisper-medium取得了最佳性能（在JASMIN上WER为5.54%，在DART上为70.37%）。它还提出了一种选择方法，能够以高精度自动识别发音正确的录音片段，从而减少人工验证的需求。