STaR: Distilling Speech Temporal Relation for Lightweight Speech Self-Supervised Learning Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2023
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910422482812928 |
|---|---|
| author | Jang, Kangwook Kim, Sungnyun Kim, Hoirin |
| author_facet | Jang, Kangwook Kim, Sungnyun Kim, Hoirin |
| contents | Albeit great performance of Transformer-based speech selfsupervised learning (SSL) models, their large parameter size and computational cost make them unfavorable to utilize. In this study, we propose to compress the speech SSL models by distilling speech temporal relation (STaR). Unlike previous works that directly match the representation for each speech frame, STaR distillation transfers temporal relation between speech frames, which is more suitable for lightweight student with limited capacity. We explore three STaR distillation objectives and select the best combination as the final STaR loss. Our model distilled from HuBERT BASE achieves an overall score of 79.8 on SUPERB benchmark, the best performance among models with up to 27 million parameters. We show that our method is applicable across different speech SSL models and maintains robust performance with further reduced parameters. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2312_09040 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | STaR: Distilling Speech Temporal Relation for Lightweight Speech Self-Supervised Learning Models Jang, Kangwook Kim, Sungnyun Kim, Hoirin Sound Computation and Language Audio and Speech Processing Albeit great performance of Transformer-based speech selfsupervised learning (SSL) models, their large parameter size and computational cost make them unfavorable to utilize. In this study, we propose to compress the speech SSL models by distilling speech temporal relation (STaR). Unlike previous works that directly match the representation for each speech frame, STaR distillation transfers temporal relation between speech frames, which is more suitable for lightweight student with limited capacity. We explore three STaR distillation objectives and select the best combination as the final STaR loss. Our model distilled from HuBERT BASE achieves an overall score of 79.8 on SUPERB benchmark, the best performance among models with up to 27 million parameters. We show that our method is applicable across different speech SSL models and maintains robust performance with further reduced parameters. |
| title | STaR: Distilling Speech Temporal Relation for Lightweight Speech Self-Supervised Learning Models |
| topic | Sound Computation and Language Audio and Speech Processing |
| url | https://arxiv.org/abs/2312.09040 |