Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913532985999360 |
|---|---|
| author | Su, Hsuan Farn, Hua Sun, Fan-Yun Chen, Shang-Tse Lee, Hung-yi |
| author_facet | Su, Hsuan Farn, Hua Sun, Fan-Yun Chen, Shang-Tse Lee, Hung-yi |
| contents | Synthetic data is widely used in speech recognition due to the availability of text-to-speech models, which facilitate adapting models to previously unseen text domains. However, existing methods suffer in performance when they fine-tune an automatic speech recognition (ASR) model on synthetic data as they suffer from the distributional shift commonly referred to as the synthetic-to-real gap. In this paper, we find that task vector arithmetic is effective at mitigating this gap. Our proposed method, SYN2REAL task vector, shows an average improvement of 10.03\% improvement in word error rate over baselines on the SLURP dataset. Additionally, we show that an average of SYN2REAL task vectors, when we have real speeches from multiple different domains, can further adapt the original ASR model to perform better on the target text domain. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_02925 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition Su, Hsuan Farn, Hua Sun, Fan-Yun Chen, Shang-Tse Lee, Hung-yi Audio and Speech Processing Artificial Intelligence Computation and Language Machine Learning Sound Synthetic data is widely used in speech recognition due to the availability of text-to-speech models, which facilitate adapting models to previously unseen text domains. However, existing methods suffer in performance when they fine-tune an automatic speech recognition (ASR) model on synthetic data as they suffer from the distributional shift commonly referred to as the synthetic-to-real gap. In this paper, we find that task vector arithmetic is effective at mitigating this gap. Our proposed method, SYN2REAL task vector, shows an average improvement of 10.03\% improvement in word error rate over baselines on the SLURP dataset. Additionally, we show that an average of SYN2REAL task vectors, when we have real speeches from multiple different domains, can further adapt the original ASR model to perform better on the target text domain. |
| title | Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition |
| topic | Audio and Speech Processing Artificial Intelligence Computation and Language Machine Learning Sound |
| url | https://arxiv.org/abs/2406.02925 |