Boosting Persian-English Speech Translation: Discrete Units & Synthetic Data

This research explores enhancements to direct speech-to-speech translation between Persian and English, a valuable contribution given the limited resources available for these language pairs. The use of discrete units and synthetic parallel data are promising approaches to improving performance, potentially benefiting wider accessibility of information.

• •Addresses challenges in translating low-resource language pairs.
• •Employs discrete units and synthetic data for enhanced performance.
• •Aims to improve direct speech-to-speech translation quality.