ACL RD-TEC 1.0 Summarization of H94-1078
Paper Title:
TECHNIQUES TO ACHIEVE AN ACCURATE REAL-TIME LARGE- VOCABULARY SPEECH RECOGNITION SYSTEM
TECHNIQUES TO ACHIEVE AN ACCURATE REAL-TIME LARGE- VOCABULARY SPEECH RECOGNITION SYSTEM
Authors: Hy Murveit and Peter Monaco and Vassilios Digalakis and John Butzberger
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- approximation
- baum-welch algorithm
- beam search
- beamsearch
- clustering
- continuous speech recognition
- decoding
- forward pass
- forward-backward search
- gaussian computation
- grammar recognition
- hidden markov
- hidden markov models
- hmm system
- hmms
- large-vocabulary continuous speech recognition
- large-vocabulary speech recognition
- lattice decoding
- modeling
- multipass search
- processor
- real-time recognition
- real-time speech recognition
- recognition
- recognition system
- recognition systems
- search
- speech recognition
- speech recognition system
- speech recognition systems
- tree search
- tree-based modeling
- vector quantization
- viterbi
- viterbi search
Other assigned terms:
- approach
- backoff
- backoff language model
- beam
- bigram
- bigram language model
- bigram model
- binary tree
- case
- cepstral vector
- cluster
- codebook
- computational complexity
- continuous speech
- distribution
- entropy
- error rate
- evaluation set
- evaluation test
- evaluations
- fact
- feature
- frame
- grammar
- hypothesis
- implementation
- language model
- language models
- large-vocabulary speech
- lattice
- lattices
- leaf
- lexicon
- markov models
- method
- names
- perplexity
- probabilities
- probability
- procedure
- process
- real-time performance
- recognition accuracy
- search strategy
- sentences
- speech corpus
- technique
- technologies
- test set
- training
- training data
- tree
- tree representation
- trees
- triphone
- uniform probability
- unigram
- unigram language model
- vocabulary
- word
- word error rate
- word error rates
- word lattices
- words
