Language Translation for Low Resource Languages

Jassiris Nunez

Co-Presenters: Individual Presentation

College: Hennings College of Science Mathematics and Technology

Major: BS.INFO/TECH

Faculty Research Mentor: Navya Martin Kollapally  

Abstract:

This project focuses on translating audio from 4 different languages into English text. We found that these existing benchmarks often fail to capture the distinctions of low-resource languages, leading to a false sense of model accuracy. Specifically using Low-Rank Adaptation (LoRA), on Whisper Medium and Whisper Turbo models. Our objective was to determine if fine-tuning could bridge the gap between automated outputs and human-verified accuracy for Spanish, Mandarin Chinese, Portuguese, and Creole.