Performance of SRI's Decipher Speech Recognition System on DARPA's CSR Task
Abstract
SRI has ported its DECIPHER(trademark) speech recognition system from DARPA's ATIS domain to DARPA's CSR domain (read and spontaneous Wall Street Journal speech). This paper describes what needed to be done to port DECIPHER(trademark), and reports experiments performed with the CSR task. The system was evaluated on the speaker-independent (SI) portion of DARPA's February 1992 "Dry-Run" WSJ0 test and achieved 17.1% word error without verbalized punctuation (NVP) and 16.6% error with verbalized punctuation (VP). In addition, we increased the amount of training data and reduced the VP error rate to 12.9%. This SI error rate (with a larger amount of training data) equaled the best 600-training-sentence speaker-dependent error rate reported for the February CSR evaluation. Finally, the system was evaluated on the VP data using microphones unknown to the system instead of the training-set's Sennheiser microphone and the error rate only increased to 26.0%.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADA457572
Entities
People
- Hy Murveit
- John Butzberger
- Mitch Weintraub
Organizations
- SRI International