ACL RD-TEC 1.0 Summarization of H90-1065
Paper Title:
TRAINING SET ISSUES IN SRI'S DECIPHER SPEECH RECOGNITION SYSTEM
TRAINING SET ISSUES IN SRI'S DECIPHER SPEECH RECOGNITION SYSTEM
Authors: Hy Murveit and Mitch Weintraub and Mike Cohen
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- continuous speech recognition
- corrective training
- database
- databases
- deleted-estimation interpolation
- digit recognition
- error rate reduction
- grammar testing
- hidden markov
- hidden markov model
- hmm-based speech recognition
- interpolation algorithm
- markov model
- model training
- modeling
- partitioning
- rate reduction
- recognition
- recognition system
- recognition systems
- recognizer
- resource management
- speaker-dependent training
- speech recognition
- speech recognition system
- training algorithm
Other assigned terms:
- acoustic models
- approach
- continuous speech
- continuous-speech
- error rate
- frame
- grammar
- hypotheses
- hypothesis
- interpolation
- large training
- perplexity
- phonological rule
- probabilities
- probability
- probability distributions
- pronunciation
- resource management task
- rule set
- sentences
- speaker-independent test set
- system description
- system evaluation
- system performance
- technique
- test set
- timit database
- training
- training data
- training set
- triphone
- vocabulary
- word
- word error rate
