Walter Lasecki and collaborators win Best Paper at W4A

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.