Comparison of Self-Supervised Speech Pre-Training Methods on Flemish Dutch
Fuente:
arXiv
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2021
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915138924183552 |
|---|---|
| author | Poncelet, Jakob Van hamme, Hugo |
| author_facet | Poncelet, Jakob Van hamme, Hugo |
| contents | Recent research in speech processing exhibits a growing interest in unsupervised and self-supervised representation learning from unlabelled data to alleviate the need for large amounts of annotated data. We investigate several popular pre-training methods and apply them to Flemish Dutch. We compare off-the-shelf English pre-trained models to models trained on an increasing amount of Flemish data. We find that the most important factors for positive transfer to downstream speech recognition tasks include a substantial amount of data and a matching pre-training domain. Ideally, we also finetune on an annotated subset in the target language. All pre-trained models improve linear phone separability in Flemish, but not all methods improve Automatic Speech Recognition. We experience superior performance with wav2vec 2.0 and we obtain a 30% WER improvement by finetuning the multilingually pre-trained XLSR-53 model on Flemish Dutch, after integration into an HMM-DNN acoustic model. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2109_14357 |
| institution | arXiv |
| publishDate | 2021 |
| record_format | arxiv |
| spellingShingle | Comparison of Self-Supervised Speech Pre-Training Methods on Flemish Dutch Poncelet, Jakob Van hamme, Hugo Audio and Speech Processing Sound Recent research in speech processing exhibits a growing interest in unsupervised and self-supervised representation learning from unlabelled data to alleviate the need for large amounts of annotated data. We investigate several popular pre-training methods and apply them to Flemish Dutch. We compare off-the-shelf English pre-trained models to models trained on an increasing amount of Flemish data. We find that the most important factors for positive transfer to downstream speech recognition tasks include a substantial amount of data and a matching pre-training domain. Ideally, we also finetune on an annotated subset in the target language. All pre-trained models improve linear phone separability in Flemish, but not all methods improve Automatic Speech Recognition. We experience superior performance with wav2vec 2.0 and we obtain a 30% WER improvement by finetuning the multilingually pre-trained XLSR-53 model on Flemish Dutch, after integration into an HMM-DNN acoustic model. |
| title | Comparison of Self-Supervised Speech Pre-Training Methods on Flemish Dutch |
| topic | Audio and Speech Processing Sound |
| url | https://arxiv.org/abs/2109.14357 |