Public Domain Speech Recognition Technology
Institute for Signal and Information Processing
Mississippi State University
Department of Electrical and Computer Engineering
phone/fax: 601-325-3149; email: email@example.com
The lack of freely available speech-to-text tools
seriously hinders research in audio information processing
technology. The Institute for Signal and Information Technology (ISIP)
has recently been funded to develop a state-of-the-art large
vocabulary conversational speech recognition system (LVCSR). This
system is being placed into the public domain, and is freely available
on the ISIP web site. The cornerstone of this system is an efficient
cross-word decoder that was recently demonstrated at ICASSP'98.
In this talk, we will review several technologies currently under
development at ISIP, including cross-word decoding in speech
recognition and Switchboard resegmentation and transcription.