Exploring Prediction Targets in Masked Pre-Training for Speech Foundation Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913655257300992 |
|---|---|
| author | Chen, Li-Wei Higuchi, Takuya Bai, He Abdelaziz, Ahmed Hussen Rudnicky, Alexander Watanabe, Shinji Likhomanenko, Tatiana Theobald, Barry-John Aldeneh, Zakaria |
| author_facet | Chen, Li-Wei Higuchi, Takuya Bai, He Abdelaziz, Ahmed Hussen Rudnicky, Alexander Watanabe, Shinji Likhomanenko, Tatiana Theobald, Barry-John Aldeneh, Zakaria |
| contents | Speech foundation models, such as HuBERT and its variants, are pre-trained on large amounts of unlabeled speech data and then used for a range of downstream tasks. These models use a masked prediction objective, where the model learns to predict information about masked input segments from the unmasked context. The choice of prediction targets in this framework impacts their performance on downstream tasks. For instance, models pre-trained with targets that capture prosody learn representations suited for speaker-related tasks, while those pre-trained with targets that capture phonetics learn representations suited for content-related tasks. Moreover, prediction targets can differ in the level of detail they capture. Models pre-trained with targets that encode fine-grained acoustic features perform better on tasks like denoising, while those pre-trained with targets focused on higher-level abstractions are more effective for content-related tasks. Despite the importance of prediction targets, the design choices that affect them have not been thoroughly studied. This work explores the design choices and their impact on downstream task performance. Our results indicate that the commonly used design choices for HuBERT can be suboptimal. We propose approaches to create more informative prediction targets and demonstrate their effectiveness through improvements across various downstream tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_10788 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Exploring Prediction Targets in Masked Pre-Training for Speech Foundation Models Chen, Li-Wei Higuchi, Takuya Bai, He Abdelaziz, Ahmed Hussen Rudnicky, Alexander Watanabe, Shinji Likhomanenko, Tatiana Theobald, Barry-John Aldeneh, Zakaria Audio and Speech Processing Sound Speech foundation models, such as HuBERT and its variants, are pre-trained on large amounts of unlabeled speech data and then used for a range of downstream tasks. These models use a masked prediction objective, where the model learns to predict information about masked input segments from the unmasked context. The choice of prediction targets in this framework impacts their performance on downstream tasks. For instance, models pre-trained with targets that capture prosody learn representations suited for speaker-related tasks, while those pre-trained with targets that capture phonetics learn representations suited for content-related tasks. Moreover, prediction targets can differ in the level of detail they capture. Models pre-trained with targets that encode fine-grained acoustic features perform better on tasks like denoising, while those pre-trained with targets focused on higher-level abstractions are more effective for content-related tasks. Despite the importance of prediction targets, the design choices that affect them have not been thoroughly studied. This work explores the design choices and their impact on downstream task performance. Our results indicate that the commonly used design choices for HuBERT can be suboptimal. We propose approaches to create more informative prediction targets and demonstrate their effectiveness through improvements across various downstream tasks. |
| title | Exploring Prediction Targets in Masked Pre-Training for Speech Foundation Models |
| topic | Audio and Speech Processing Sound |
| url | https://arxiv.org/abs/2409.10788 |