Adaptability of ASR Models on Low-Resource Language: A Comparative Study of Whisper and Wav2Vec-BERT on Bangla

TL;DR

This study compares Whisper and Wav2Vec-BERT ASR models on Bangla, revealing Wav2Vec-BERT's superior performance and efficiency in low-resource language settings.

Contribution

It provides a systematic comparison of two leading ASR models on Bangla, highlighting Wav2Vec-BERT's advantages in accuracy and resource utilization.

Findings

Wav2Vec-BERT outperforms Whisper in WER and CER

Wav2Vec-BERT requires fewer computational resources

Systematic hyperparameter tuning improves model performance

Abstract

In recent years, neural models trained on large multilingual text and speech datasets have shown great potential for supporting low-resource languages. This study investigates the performances of two state-of-the-art Automatic Speech Recognition (ASR) models, OpenAI's Whisper (Small & Large-V2) and Facebook's Wav2Vec-BERT on Bangla, a low-resource language. We have conducted experiments using two publicly available datasets: Mozilla Common Voice-17 and OpenSLR to evaluate model performances. Through systematic fine-tuning and hyperparameter optimization, including learning rate, epochs, and model checkpoint selection, we have compared the models based on Word Error Rate (WER), Character Error Rate (CER), Training Time, and Computational Efficiency. The Wav2Vec-BERT model outperformed Whisper across all key evaluation metrics, demonstrated superior performance while requiring fewer computational resources, and offered valuable insights to develop robust speech recognition systems in low-resource linguistic settings.