Continued Pretraining for Domain Adaptation of Wav2vec2.0 in Automatic Speech Recognition for Elementary Math Classroom Settings
This work addresses the problem of creating robust ASR systems for classroom settings to aid teachers and students, but it is incremental as it applies an existing adaptation method to a specific domain.
The paper tackled adapting Wav2vec2.0 for automatic speech recognition in elementary math classrooms using continued pretraining, resulting in a reduction of Word Error Rate by over 10% and improved robustness to various classroom conditions and demographics.
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.