[Faculty of Modern Social Studies] Students present research results at international conference

KHINE ZAR THWE, a third-year student in Faculty of Modern Social Studies, participated in the 2026 RISP International Workshop on Nonlinear Circuits, Communications and Signal Processing (NCSP'26), an international conference held in Hawaii from February 27 to March 2, and presented his research results online.

Presentation title
How much data is enough? Fine-tuning large pretrained ASR models for low-resource Burmese language

Content
While speech recognition technology has achieved high accuracy in major languages such as English, improving performance in low-resource languages with limited data remains a major challenge. In this presentation, we conducted experiments using large-scale pre-trained speech recognition models (whisper, wav2vec Bert 2.0, XLS-R) on Burmese, one of the low-resource languages, and presented the results of our analysis of the possibility of achieving practical recognition accuracy and the amount of training data required to achieve it.

Student comments
I am very pleased to have had the opportunity to present my research on Burmese automatic speech recognition (ASR) at this conference. It was a very meaningful opportunity to share my research and receive valuable feedback from other researchers. Although I was a little nervous about presenting in front of such excellent researchers from other universities, it was a very stimulating and encouraging experience.

(Source: Faculty of Modern Social Studies and Center for International Exchange and Study Abroad)