Continued Pretraining for Domain Adaptation of Wav2vec2.0 in Automatic Speech Recognition for Elementary Math Classroom Settings
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_ | 1866913358320500736 |
|---|---|
| author | Attia, Ahmed Adel Demszky, Dorottya Ogunremi, Tolulope Liu, Jing Espy-Wilson, Carol |
| author_facet | Attia, Ahmed Adel Demszky, Dorottya Ogunremi, Tolulope Liu, Jing Espy-Wilson, Carol |
| contents | Creating Automatic Speech Recognition (ASR) systems that are robust and resilient to classroom conditions is paramount to the development of AI tools to aid teachers and students. In this work, we study the efficacy of continued pretraining (CPT) in adapting Wav2vec2.0 to the classroom domain. We show that CPT is a powerful tool in that regard and reduces the Word Error Rate (WER) of Wav2vec2.0-based models by upwards of 10%. More specifically, CPT improves the model's robustness to different noises, microphones, classroom conditions as well as classroom demographics. Our CPT models show improved ability to generalize to different demographics unseen in the labeled finetuning data. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_13018 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Continued Pretraining for Domain Adaptation of Wav2vec2.0 in Automatic Speech Recognition for Elementary Math Classroom Settings Attia, Ahmed Adel Demszky, Dorottya Ogunremi, Tolulope Liu, Jing Espy-Wilson, Carol Computation and Language Artificial Intelligence Audio and Speech Processing Creating Automatic Speech Recognition (ASR) systems that are robust and resilient to classroom conditions is paramount to the development of AI tools to aid teachers and students. In this work, we study the efficacy of continued pretraining (CPT) in adapting Wav2vec2.0 to the classroom domain. We show that CPT is a powerful tool in that regard and reduces the Word Error Rate (WER) of Wav2vec2.0-based models by upwards of 10%. More specifically, CPT improves the model's robustness to different noises, microphones, classroom conditions as well as classroom demographics. Our CPT models show improved ability to generalize to different demographics unseen in the labeled finetuning data. |
| title | Continued Pretraining for Domain Adaptation of Wav2vec2.0 in Automatic Speech Recognition for Elementary Math Classroom Settings |
| topic | Computation and Language Artificial Intelligence Audio and Speech Processing |
| url | https://arxiv.org/abs/2405.13018 |