The paper explores how automated speech recognition and crowd-sourced human correction and generation of transcripts can be traded off to improve accuracy and latency.
The paper explores how automated speech recognition and crowd-sourced human correction and generation of transcripts can be traded off to improve accuracy and latency.