CMU Professor’s Work To Bridge Language Gaps Using Tech Earns National Recognition
There are 7,000 languages spoken across the world, and it is Graham Neubig’s mission to get computers to better process them all. The Carnegie Mellon University professor is studying how to make voice-controlled assistants like Alexa, Siri, and Google Translate proficient in more languages.
Currently, Siri is only available in 21 languages. Google Translate supports 103 languages, but its speech-to-text feature is only functional in 41.
“For most of the languages in the world, it’s not supported,” Neubig said. “Even languages that have millions of speakers, like many of the languages in Africa, central Asia [and] Eastern Europe.”
To get these assistants to speak and translate more languages, their algorithms need more data, according to Neubig. A lot of data has been translated from English to another language, which is why it’s so widely available.
Neubig studies how to improve language processing, the technology that serves as the foundation for virtual assistants, instant translation tools and autocomplete functions in text messaging apps. The more languages this technology can serve, the easier it will be for people to communicate with each other —and with their virtual assistants— in their native tongues.
As access to smart devices continues to expand across the globe, this technology could too, according to Neubig. That could have big implications for developing countries.
“Even in places where people don’t read and write most of the time,” he said. “I think that voice technology, speech recognition, assistants that you can speak to and get information from without having to go through typing or looking at your computer screen are very good" for communicating across languages.
Neubig also researches how to make language processing done by a computer as nuanced as when it’s translated by a human. Many languages don’t have one-to-one translations for phrases and different cultural implications based on semantics and tone.
“In Japanese, it’s a lot stronger of a statement to say ‘No. I can’t do that,” and instead you would say, ‘No. That is difficult,’” said Neubig, who speaks both English and Japanese. “There’s a little bit more softening of how you say things [in Japanese] ... inferring the other person’s mentality.”
Enhancing language processing technology can allow people to fully express themselves through translation apps or best direct their virtual assistants, according to Neubig.
“I really like the idea of this technology as something that allows people to speak in their own words,” instead of being forced to homogenize syntax or tone into a language the computer can understand. “This technology is going to help us globalize in a way that still respects the uniqueness of cultures all over the globe.”
Neubig’s work earned national recognition last week: he was named a finalist for a Blavatnik Award for Young Scientists in the physical sciences and engineering category. The Blavatnik Family Foundation awards three young scientists with a $250,000 prize each year. The three winners will be announced next month.