ACL RD-TEC 1.0 Summarization of H89-1011
Paper Title:
SPEAKER ADAPTATION FROM LIMITED TRAINING IN THE BBN BYBLOS SPEECH RECOGNITION SYSTEM
SPEAKER ADAPTATION FROM LIMITED TRAINING IN THE BBN BYBLOS SPEECH RECOGNITION SYSTEM
Authors: Francis Kubala and Ming-Whel Feng and John Makhoul and Richard Schwartz
Primarily assigned technology terms:
- algorithm
- approximation
- byblos speech recognition
- class labeling
- computing
- continuous speech recognition
- database
- dtw-based speaker adaptation
- hidden markov
- hidden markov model
- linear regression
- markov model
- matching
- modeling
- normalization
- normalization procedure
- parsing
- rapid speaker adaptation
- recognition
- recognition system
- recognition systems
- regression
- resource management
- smoothing
- smoothing problem
- speaker adaptation
- speaker-dependent training
- speech recognition
- speech recognition system
- training procedure
Other assigned terms:
- approach
- byblos system
- cluster
- co-occurrence
- codebook
- conditional probability
- continuous speech
- convergence
- density function
- derivational equivalence
- error rate
- estimation
- evaluation test
- experimental results
- fact
- frame
- grammar
- grammars
- hmm model
- labeling
- lexicon
- many-to-many mapping
- mapping
- maps
- method
- nist
- performance evaluation
- perplexity
- phoneme
- phonemes
- probabilities
- probability
- probability density
- probability density function
- procedure
- sentence
- sentences
- sparse data
- speech data
- technology
- test data
- test set
- text
- tokens
- training
- training and test data
- training data
- training material
- training set
- transformation
- transition probabilities
- word
- word error rate
- word error rates
- word sequence
- word-pair grammar
