Brigham Young University researchers using tech to preserve languages, improve translation

SALT LAKE CITY — These days, language seems more accessible than ever with learning and translation apps quite literally at our fingertips.

Still, thousands of languages remain mostly invisible to modern translation tools.

According to Ethnologue, about 7,170 known languages — at varying levels, of course — are currently spoken around the world. For comparison, Duolingo offers 42 languages, and Google Translate includes nearly 250 languages.

The gap exists, and it's one students and researchers at Brigham Young University are helping to close through a project called Pathsay.

Through Pathsay, students in the school's MATRIX (machine translation research and interlingual experimentation) Lab are partnering with international BYU-Pathway Worldwide students to collect speech and text data for low-resource languages, "helping preserve linguistic heritage and improve access to translation tools for communities often overlooked by mainstream technology," according to a release from the school.

Students in the MATRIX lab use the collected data to create AI models that can then recognize, translate and generate the low-resource languages of the world.

Already, the project has collected over 750,000 recorded sentences and 2,200 hours of audio across 30 different languages, including Afrikaans, Bislama, Chichewa, Cusco Quechua, Dholuo, Efik, Fante and Fon, among others. By the end of the year, researchers hope to include 50 languages.

The release from the university also noted that of the some 7,000 languages spoken today, fewer than 50 have enough digital data available for "truly useful translation tools," with only the top dozen languages like English, Spanish and Chinese considered "high-resource."

"Any language beyond the top 100 is what we would call low-resource," BYU computer science professor Steve Richardson said in a statement. "There are not enough translated sentences online to train models like Google Translate to handle those languages. Then, the other half of the world's languages aren't even written, and you find what we call endangered languages — languages that may just disappear with a few dozen native speakers left. For those kinds of languages, the goal is to preserve the culture and the heritage — to not lose the valuable knowledge of those people."

And it's not just obscure languages that aren't supported by mainstream tech. Far from it, actually, with the release noting that many of the languages overlooked are actively spoken by hundreds of thousands or even millions of people.

Pathway program studentsalso get the chance to put their own, personal touches on the project, with students from countries such as South Africa, Nigeria and Kenya recording sentences in their native language through a program developed by students in the lab.

"I get to preserve my own language," Thomas, a BYU Pathway student from Ghana, said in a Y Magazine podcast interview. "People do not really get to speak their local languages even to their colleagues, and so getting a chance to learn ... (more) about your local language has been a really incredible thing to me."

Related topics