Abstract
From the background information supplied by the inventors, news correspondents o btained the following quote: “Automatic speech recognition (ASR) often employs n eural networks (and/or machine learning techniques). Such networks must be train ed on samples of speech audio with transcriptions checked by humans. Supervised machine learning requires labeled data. Checking transcriptions is part of label ing data for training automatic speech recognition using machine learning. Label ing data has a fairly low skill requirement and can be done at any time of day. As a result, it is a perfect task for people who work remotely. Many times, this transcription is done by part-time employees or non-employee contractors, who l isten to and transcribe recordings of human speech. Other times, humans check an d confirm machine generated transcriptions of speech.
NETL
NSTL