A team at Brigham Young University (BYU) is working to bridge the digital divide for thousands of languages that currently lack sufficient data for modern translation tools. Through the Pathsay project, students in the university's MATRIX lab are collaborating with international BYU-Pathway Worldwide students to gather text and audio recordings for low-resource languages.

While there are approximately 7,000 languages globally, fewer than 50 possess the digital data required to develop effective translation software. Most mainstream tools focus on high-resource languages like English, Chinese, and Portuguese. BYU computer science professor Steve Richardson noted that languages outside the top 100 often lack the online translated content necessary to train models like Google Translate, leaving many communities—including those with millions of speakers—without adequate technological support.

To date, the Pathsay project has compiled over 750,000 sentences and 2,200 hours of audio across 30 languages, including Efik, Fante, Bislama, and Cusco Quechua. The team aims to increase this coverage to 50 languages by the end of the year. Participants from countries such as Nigeria, Kenya, and South Africa record native speech, a process that students say helps preserve their linguistic heritage.

Beyond linguistic preservation, the project is driven by a goal to enable seamless communication between speakers of different languages, including those that are unwritten or endangered. For the research team, the effort also carries a spiritual objective: making religious resources accessible to global populations in their native tongues. By utilizing the gathered data, the MATRIX lab is developing AI models capable of recognizing, translating, and generating these underrepresented languages.

Source: Phys.org