It’s easy to take the overhead announcements in train stations and airports for granted, but oftentimes they present passengers with indispensable information about weather conditions, the status of their trip, and more. Of course, it can get a bit dry listening to the same, monotonous voice drone on and on about security measures and whatnot, but when you’re running half an hour late for your flight, there are few things more comforting than hearing someone over the intercom announce that your flight’s been delayed by just enough time for you to make it on board.
On the other hand, individuals who are deaf or hard-of-hearing don’t get to take these announcements for granted. In an effort to break down this language barrier for people with hearing loss, researchers at the University of Amsterdam have begun developing a machine translation (MT) tool for signed languages, that can help deaf and hard-of-hearing individuals communicate with hearing individuals in scenarios where it may not always be practical to have an interpreter present at all times.
Floris Roelofson, an associate professor at the university, founded SignLab Amsterdam in 2020 alongside other researchers at the university to explore sign linguistics, or the structure of signed languages. The group initially began the aforementioned MT project in response to COVID-19, in an effort to “improve communications between healthcare professionals and deaf patients,” however the group plans to apply this research to other scenarios as well, including train stations and airports.
“Governments, companies, and anyone providing services should take (deaf or hard-of-hearing) passengers into account, certainly governments,” Roelofson said in an interview with the online magazine Railway Technology. “If they’re providing a service that is paid out of tax money that is also paid by deaf people, then they should have equal access to those services.”
SignLab Amsterdam’s MT system utilizes an animated avatar which translates written language into signed language. Currently, MT for signed languages lags behind MT for spoken languages, in part because artificial intelligence research regarding signed languages often disregards natural language processing methods (which are specialized for dealing with the complex syntax and semantics of human language), instead drawing on research from the field of computer vision.
“Scientific investigation of sign languages is not only necessary to obtain a general understanding of human languages, but also to diminish the communication barrier between deaf and hearing people,” reads SignLab Amsterdam’s website.