Real-time Transcription through Crowdsourcing

Scribe is a research project led by researcher Jeff Bigham that collects audio from a mobile device and schedules workers to perform captioning on small audio segments. The distributed captions are then merged into a single text, returning a consolidated transcript to the mobile user.

We worked with Adobe Air on the mobile client to stream audio to an Adobe Media Server and with javascript and AJAX in the front-end worker application. Back-end development included creating diagnostic and debugging tools in a LAMP Stack as well as integrating with the java process in charge of merging the distributed transcriptions.

Go back to Projects