ISCA International Workshop on Speech and Language Technology in Education (SLaTE 2009)

Wroxall Abbey Estate, Warwickshire, England
September 3-5, 2009

Voice Race and Voice Scatter: Online Educational Games for Collecting Orthographically-Labeled Speech Data

Alexander Gruenstein, Ian McGraw, Andrew Sutherland

MIT, Cambridge, MA, USA

Voice Race and Voice Scatter are online education games available on the popular flashcard website Quizlet.com. Quizlet users can make and share sets of virtual flashcards, which each contain a term on one side and a definition on the other. Quizlet boasts 420,000 registered users who have created over 875,000 sets of flashcards, which altogether contain more than 24 million individual flashcards.
   Voice Race and Voice Scatter use the publicly available WAMI Javascript API, which makes it easy to incorporate speech recognition capabilities into Web applications, to provide a fun way for users to study flashcards on the website by speaking. Moreover, by using recognition confidence scores and contextual information from the games, it is possible to automatically orthographically label a large portion of the collected utterances with near-human accuracy. As such, the games provide diversion, educational value, and labeled speech data.

Bibliographic reference.  Gruenstein, Alexander / McGraw, Ian / Sutherland, Andrew (2009): "Voice race and voice scatter: online educational games for collecting orthographically-labeled speech data", In SLaTE-2009 (abstract).