Hiring a professional captioning service to help students with hearing disabilities get information presented in class or at events requires a lot of advance planning and can be costly. Automatic speech recognition programs, while quick and convenient, have unusually high error rates. However, a program called Scribe, developed by Walter Lasecki, assistant professor of information and of computer science and engineering, algorithmically combines the efforts of several non-professional translators (such as peers or work-study students) to form accurate, real-time captions in less than four seconds.
“What we did is tried to essentially democratize this process,” said Lasecki, who began the work at the University of Rochester. “By doing turn-taking and then aggregation, we can actually get a much more reliable signal.” Lasecki said there is still room for improvement, but he hopes one day the program can be helpful to students with hearing disabilities and university offices that assist them.
