Arabic ASR on the SADA Large-Scale Arabic Speech Corpus with Transformer-Based Models
This work addresses speech recognition for Arabic speakers, particularly in noisy and dialect-rich environments, but it is incremental as it applies existing methods to a new dataset.
The researchers tackled Arabic automatic speech recognition by evaluating state-of-the-art models on the large-scale SADA dataset, finding that the MMS 1B model fine-tuned with a language model achieved a word error rate of 40.9% and character error rate of 17.6% on the clean test set.
We explore the performance of several state-of-the-art automatic speech recognition (ASR) models on a large-scale Arabic speech dataset, the SADA (Saudi Audio Dataset for Arabic), which contains 668 hours of high-quality audio from Saudi television shows. The dataset includes multiple dialects and environments, specifically a noisy subset that makes it particularly challenging for ASR. We evaluate the performance of the models on the SADA test set, and we explore the impact of fine-tuning, language models, as well as noise and denoising on their performance. We find that the best performing model is the MMS 1B model finetuned on SADA with a 4-gram language model that achieves a WER of 40.9\% and a CER of 17.6\% on the SADA test clean set.