EARS: Rich Transcription
Rich Transcription
We are working to generate readable transcriptions of conversational speech
in multiple languages. "Readable" here means incorporating capitalization,
punctuation, and speaker markers; but it also means making major improvements
in speech recognition performance, since word errors are still significant
in this type of task.
OBJECTIVES
TEAM (site leaders)
SRI (lead site) - Andreas Stolcke
ICSI - Barbara Peskin
UW - Mari Ostendorf
TASKS
The main components can be broken down into four tasks:
(with Principal Investigators listed first, co-PIs second)
Core Automatic Speech Recognition (ASR) Innovation
Horacio Franco -SRI
Rapid Development of ASR in New Languages
and Domains (Portability)
Mari Ostendorf -UW, Kristin Precoda -SRI
Metadata Extraction and Modeling
Barbara Peskin -ICSI, Elizabeth Shriberg -SRI
Evaluation System Development
V.R.R. Gadde -SRI
Back to the EARS main page