Which Data Matter? Embedding-Based Data Selection for Speech Recognition
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910051357163520 |
|---|---|
| author | Aldeneh, Zakaria Seto, Skyler de Seyssel, Maureen Chi, Jie Gu, Zijin Higuchi, Takuya Jung, Jee-weon Watanabe, Shinji Grangier, David Theobald, Barry-John Likhomanenko, Tatiana |
| author_facet | Aldeneh, Zakaria Seto, Skyler de Seyssel, Maureen Chi, Jie Gu, Zijin Higuchi, Takuya Jung, Jee-weon Watanabe, Shinji Grangier, David Theobald, Barry-John Likhomanenko, Tatiana |
| contents | Modern ASR systems are typically trained on large-scale pseudo-labeled, in-the-wild data spanning multiple domains. While such heterogeneous data benefit generalist models designed for broad deployment, they pose challenges for specialist models targeting specific domains: specialist models lack the capacity to learn from all available data, and one must pay closer attention to addressing the mismatch between training and test conditions. In this work, we study targeted data selection as a strategy to address these challenges, selecting relevant subsets from 100k hours of in-the-wild training data to optimize performance on target domains. We represent speech samples using embeddings that capture complementary characteristic--speaker attributes, phonetic content, and semantic meaning--and analyze how relevance and diversity along these axes when performing data selection affect downstream ASR performance. Our experiments with CTC-based Conformer models show that training on a strategically selected 5% subset can exceed the performance of models trained on the full dataset by up to 36.8% relative WER reduction on target domains. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_05819 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Which Data Matter? Embedding-Based Data Selection for Speech Recognition Aldeneh, Zakaria Seto, Skyler de Seyssel, Maureen Chi, Jie Gu, Zijin Higuchi, Takuya Jung, Jee-weon Watanabe, Shinji Grangier, David Theobald, Barry-John Likhomanenko, Tatiana Sound Modern ASR systems are typically trained on large-scale pseudo-labeled, in-the-wild data spanning multiple domains. While such heterogeneous data benefit generalist models designed for broad deployment, they pose challenges for specialist models targeting specific domains: specialist models lack the capacity to learn from all available data, and one must pay closer attention to addressing the mismatch between training and test conditions. In this work, we study targeted data selection as a strategy to address these challenges, selecting relevant subsets from 100k hours of in-the-wild training data to optimize performance on target domains. We represent speech samples using embeddings that capture complementary characteristic--speaker attributes, phonetic content, and semantic meaning--and analyze how relevance and diversity along these axes when performing data selection affect downstream ASR performance. Our experiments with CTC-based Conformer models show that training on a strategically selected 5% subset can exceed the performance of models trained on the full dataset by up to 36.8% relative WER reduction on target domains. |
| title | Which Data Matter? Embedding-Based Data Selection for Speech Recognition |
| topic | Sound |
| url | https://arxiv.org/abs/2603.05819 |