Network of volunteer translators wants to make refugee languages ​​more accessible. Will AI help?

NEW YORK (AP) — They could be Tigrinya speakers fleeing Eritrea’s authoritarian government has introduced unlimited military service politics. Or Rohingya people fleeing ethnic violence in Myanmar. But refugees navigating resettlement often face a common obstacle: poor machine translation and a shortage of translators familiar with their lesser-served languages.

Tarjimly, a Google-backed nonprofit that’s been described as “Uber for translators,” aims to help asylum seekers overcome that hurdle. With a new AI partnership, Tarjimly is training beyond large language models while allowing its volunteers to respond more urgently to translators’ needs. It’s a feedback loop in which humans teach machines the nuances of each language, sharing data from individual conversations and improving the automatic translations.

And it’s this uniquely human domain of language that Tarjimly co-founder Atif Javed says exemplifies the difficult balance between individual ingenuity and technological progress. It’s the personal connection he needs that shows why AI’s rapid advances shouldn’t generally be a cause for concern.

Languages ​​popular in Global South — such as Dari and Pashto, widely spoken in Afghanistan, where one of the world’s largest protracted refugee crises is underway — have the worst quality of coverage, according to Javed. He believes he is well-positioned to supplement the English-dominated online information stores that train services like Google Translate with the more diverse data sets of his mobile app.

Tarjimly connects refugees with on-demand interpreters who can communicate during meetings with social workers, immigration officials and doctors, and records the meetings for AI training. To comply with patient privacy laws, Tarjimly anonymizes conversations in its app. Javed said the nonprofit also has an option for “no-log” sessions, in which no data is stored for alternative uses.

According to Javed, many of the 60,000 volunteers are multilingual refugees who understand not only the native language of those they are speaking, but also the crisis that brought them there.

Among them is Roza Tesfazion, a 26-year-old refugee from Eritrea who works professionally as a translator for the UK government. Fluent in Amharic and Tigrinya, she learned English and Swahili to help her immigrant family overcome language barriers when they first moved to Kenya.

Tesfazion said she translates for free because she knows “how emotional it is” for the people on the other end of her sessions.

“You have to have a little bit of human emotion in you,” she said.

Tarjimly’s founders say the sensitive nature of their mission lends itself better to nonprofit status than a corporate structure. Users come from highly precarious situations, and the nonprofit works with established humanitarian groups, including Catholic Charities, the International Rescue Committee and the United Nations International Organization for Migration.

According to Javed, the job requires a level of trust that would be hard to come by in a “profit-driven, competitive world.” “The fundamental driver of our success is the community we’ve built.”

But this community also has room for AI. A $1.3 million grant from Google.org has enabled the creation of “First Pass,” a tool that instantly generates translations for human volunteers to correct. The new information center will share its language data with partners, including Google, in early 2025.

But according to Ranjit Singh, a researcher at Data & Society, improving a more diverse language library will require conversational data on a much larger scale than Tarjimly can provide on its own.

Singh, who studies the social implications of automation and inclusive digital solutions, said translation services will always need a “real middleman.”

“One part is translating, and the other part is trying to understand someone’s life situation,” he said. “Technology helps us do some of that work. But at the same time, it’s also quite social.”

Tarjimly was inspired by the time Javed spent volunteering with Arabic-speaking people in refugee camps in Greece and Turkey after graduating from the Massachusetts Institute of Technology and working in Silicon Valley. Javed, a Muslim American whose family immigrated to the United States in 2001, said it reminded him of his own childhood, when he translated for his refugee grandmother.

His life experience is one reason Elevate Prize Foundation CEO Carolina Garcìa Jayaram said her organization awarded Tarjimly $300,000 last year. That “close leadership” helps nonprofits better understand developments like artificial intelligence, which “can be a cause for both excitement and anxiety,” Jayaram said. The risk-averse philanthropic sector can be slow to keep up with disruptive new technologies, she noted, but it shouldn’t ignore their positive applications.

“This is a great example of how not to get stuck in the AI ​​scare complex,” she said. “Go to the leaders who are closest to these problems and say, ‘How will AI open up possibilities and opportunities for your organization?’”

—-

Associated Press coverage of philanthropy and nonprofits is supported by AP’s partnership with The Conversation US, with funding from Lilly Endowment Inc. The AP is solely responsible for this content. For all of AP’s coverage of philanthropy, visit https://apnews.com/hub/philanthropy.