Has anyone worked on streaming or online approach with DeepSpeech?

(Btofel) #1

Curious if there is already work in the public domain on this. Online in this sense I mean to mean that some component of the model is altered in real-time to account for speech heard just before. Or, if the model isn’t changed, some techniques are employed to work with the audio as a continuous stream, realizing topic of conversation and supplying back-fixes to previous transcription.