AI meets ancient text: UP team develops Baybayin translation tool

Development and Purpose

A team from the University of the Philippines, led by mathematician Renier Mendoza, Rachelle Sambayan, and student Rodney Pino, created an AI-powered tool to translate Baybayin. They identified a significant lack of digital tools for the precolonial script compared to other major Asian languages and sought to bridge this gap to help historians and scholars decode ancient texts.

Technology and Methodology

The tool uses machine-learning algorithms and a support vector machine (SVM) to recognize Baybayin characters. The team built a database of 17,000 images, encompassing 1,000 variations for each of the 17 Baybayin characters, to train the software to distinguish the script from Latin characters.

Current Capabilities and Future Goals

While the software can process full paragraphs, it is in the early stages and faces challenges with vowel disambiguation and contextual word choice. The researchers are working to improve its speed and linguistic accuracy, with plans to develop a mobile version. Currently, the team has made their Baybayin database public to encourage further academic research.