Achieving Acceptable Accuracy in a Low-Cost, Assistive Note-Taking, Speech Transcription System

T. Way, R. Kheir, and L. Bevilacqua (USA)


Real-time speech recognition, assistive note-taking, accurate-enough transcription, low-cost assistive technology.


Recent advancements in speech recognition technology and the widespread availability of inexpensive computers mean that real-time transcription in the classroom is becoming practical. This paper reports on the accuracy and readability achieved using a low-cost speech transcription system that assists deaf and hard-of-hearing students with note-taking. The design is presented for a system that attempts to balance issues of availability and cost with usability and accuracy to provide a viable and fully automatic alternative to a human signer or note taker. While this study focuses on empirical measures of design choices and accuracy, anecdotal results are also presented, with particular attention paid to the practical trade-offs between cost, accuracy and readability of a real-time transcription system.

Important Links:

Go Back