ACL RD-TEC 1.0 Summarization of H94-1063
Paper Title:
HIGH-ACCURACY LARGE-VOCABULARY SPEECH RECOGNITION USING MIXTURE TYING AND CONSISTENCY MODELING
HIGH-ACCURACY LARGE-VOCABULARY SPEECH RECOGNITION USING MIXTURE TYING AND CONSISTENCY MODELING
Authors: Vassilios Digalakis and Hy Murveit
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- approximation
- bootstrap
- clustering
- consistency modeling
- continuous speech recognition
- database
- decoding
- discriminant analysis
- forward-backward algorithm
- genonic hmm
- hidden markov
- hidden markov models
- hmm recognition
- hmm system
- hmm-based speech recognizer
- hmms
- large-vocabulary continuous speech recognition
- large-vocabulary speech recognition
- linear discriminant
- linear discriminant analysis
- modeling
- progressive-search
- pruning
- recognition
- recognition system
- recognizer
- reestimation
- search
- smoothing
- speech recognition
- speech recognition system
- speech recognizer
- splitting
- viterbi
- viterbi decoding
Other assigned terms:
- approach
- bigram
- case
- cluster
- clustering procedure
- codebook
- coefficient
- continuous speech
- correlation
- development set
- distribution
- duration
- entropy
- error rate
- estimation
- evaluation set
- experimental results
- feature
- frame
- independence assumption
- index
- knowledge
- language model
- language models
- large-vocabulary speech
- lattices
- likelihood
- mapping
- markov models
- mutual information
- names
- opinions
- probabilities
- procedure
- recognition task
- search space
- seed
- sentences
- speech recognition performance
- system development
- tags
- term
- training
- training data
- transformation
- trigram
- trigram language model
- vocabulary
- vocal tract
- word
- word error rate
- word lattices
- wsj corpus
- wsj development set
