I spoke about the following ideas during BeyondBitcoin episode 200, give it a listen if you're interested! RSVP thread. Raw recording
BeyondBitcoin audio transcription!
Google charges $0.006 per 15secs ($260 for 3hrs audio), could an open source alternative to Google cloud API prove cheaper & perhaps able to undercut Google?
Possible (FOSS) software: CMU Sphinx, Kaldi, Julius, VoxForge.
Alternatively to an automated audio transcription mechanism, you could implement a mechanical turk system in which users transcribe the audio manually and in return users gain credit (to which we would base GRC rewards upon).
I began to translate the interview with Travis Desell on the YouTube channel
If more people would do something like that I think more people could be interested in these hangouts. Sometimes it is really difficult to unterstand the conference.
Excellent! Thanks for taking time to translate the interviews for the Gridcoin community, what language are you translating them to?
I try to translate in German. I'm German native Speaker, but often it is difficult to unterstand the speech and then sometimes the english transcription helps. If someone could improve the english transcription before I translate it into German, it would be very helpful.
Maybe it would be helpful, if we use the autmatically created transcription from google in the YouTube channel and improve the text manualy. The next step would be to translate this text into diffrent languages.
Yeah, this is indeed a manual process that can be performed - extracting the auto generated transcription from youtube videos once they're processed. There is a delay before google performs this transcription though.
We could do some kind of mechanical turk system for watching chunks of the video and confirming the extracted subtitles.
We could simply use google translate against these extracted transcriptions, but that would be quite messy.