ACL RD-TEC 1.0 Summarization of H90-1062
Paper Title:
IMPROVED ACOUSTIC MODELING FOR CONTINUOUS SPEECH RECOGNITION
IMPROVED ACOUSTIC MODELING FOR CONTINUOUS SPEECH RECOGNITION
Authors: C.-H. Lee and E. Giachin and L. R. Rabiner and R. Pieraccini and A. E. Rosenberg
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- analyzer
- atis
- automatic feature selection
- beam search
- beam search algorithm
- clustering
- continuous density hmm
- continuous speech recognition
- corrective training
- database
- decoder
- decoding
- error rate reduction
- error reduction
- feature analysis
- feature selection
- feature selection algorithm
- finite state
- flame-synchronous beam search
- frame-synchronous beam search
- hidden markov
- hidden markov model
- information system
- k-means
- k-means clustering
- k-means training
- language analyzer
- language decoder
- large vocabulary speech recognition
- lexical modeling
- markov model
- modeling
- normalization
- optimization
- pattern recognition
- phonetic transcription
- post-processing
- principal component analysis
- rate reduction
- recognition
- recognition procedure
- recognition system
- recognition systems
- recognizer
- resource management
- sampling
- search
- search algorithm
- segmentation
- selection algorithm
- selection strategy
- speaker verification
- speaker-independent recognition
- spectral analysis
- speech decoding
- speech recognition
- speech recognition system
- speech recognition systems
- training algorithm
- training procedure
- transcription
- unit selection
- word discrimination
- word recognition
Other assigned terms:
- 109-speaker training set
- approach
- beam
- byblos system
- case
- coefficient
- content words
- continuous speech
- darpa naval resource management
- delta energy
- dictionary
- error rate
- estimation
- evaluation data
- evaluations
- experimental results
- fact
- feature
- feature set
- feature vector
- feature vectors
- finite state network
- frame
- function words
- gaussian mixture
- grammar
- histogram
- hypotheses
- implementation
- khz rate
- language model
- large vocabulary speech
- lattice
- lexical entry
- lexical representation
- lexicon
- likelihood
- linear combination
- linguistic
- mapping
- methodology
- model parameters
- perplexity
- phonemes
- phonetic representation
- phonological rules
- priori
- probabilities
- procedure
- pronunciation
- recognition accuracy
- recognition errors
- recognition task
- resource management task
- segments
- selection rule
- semantic
- semantic rules
- sentence
- sentences
- silence
- speech database
- speech recognition performance
- speech recognition task
- spoken language
- symbol
- term
- terms
- test data
- test set
- theory
- training
- training corpus
- training data
- training material
- training set
- transformation
- transition probabilities
- unit separation distance
- utterance
- vocabulary
- word
- word boundaries
- word lattice
- word model
- word pair
- word pair grammar
- word sequence
- word-pair grammar
- words
