Language models can fix speech-to-text errors for rare languages

Original: Can Large Language Models Reliably Correct Errors in Low-Resource ASR? A Contamination-Aware Case Study on West Frisian

Source: arxiv.org ↗