ACL RD-TEC 1.0 Summarization of H93-1016
Paper Title:
AN OVERVIEW OF THE SPHINX-II SPEECH RECOGNITION SYSTEM
AN OVERVIEW OF THE SPHINX-II SPEECH RECOGNITION SYSTEM
Authors: Xuedong Huang and Fileno Alleva and Mei-Yuh Hwang and Ronald Rosenfeld
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- backoff bigram
- baum-welch reestimation
- beam search
- classification
- clustering
- continuous speech recognition
- cross validation
- database
- database construction
- decision tree
- decision-tree
- decoding
- discriminative training
- error rate reduction
- error reduction
- feature extraction
- hidden markov
- hidden markov model
- hidden markov models
- hill-climbing
- hmm-based speech recognition
- hmms
- language modeling
- language training
- likelihood estimation
- markov model
- maximum likelihood
- maximum likelihood estimation
- model optimization
- modeling
- multi-pass rescoring
- multi-pass search
- n-best paradigm
- n-best paradigm \
- normalization
- optimization
- pronunciation optimization
- rate reduction
- recognition
- recognition system
- recognition systems
- reestimation
- rescoring
- search
- search algorithm
- searching
- senonic modeling
- sentence classification
- sentence recognition
- smoothing
- speaker clustering
- speech recognition
- speech recognition system
- speech recognition systems
- spelling
- sphinx-ii
- stochastic language modeling
- stress test evaluation
- table lookup
- validation
- viterbi
- viterbi beam
- viterbi beam search
Other assigned terms:
- acoustic models
- approach
- backoff
- beam
- bigram
- bigram language model
- classification error
- codebook
- computational complexity
- continuous speech
- data set
- density function
- dimensionality
- distance measure
- distribution
- duration
- error rate
- estimation
- evaluation function
- evaluations
- feature
- feature set
- feature vector
- frame
- hypotheses
- hypothesis
- implementation
- knowledge
- language information
- language model
- language models
- lattice
- leaf
- likelihood
- linear combination
- linguistic
- linguistic information
- linguistic knowledge
- markov models
- measure
- measures
- method
- model parameters
- modeling power
- n-gram
- noise
- priori
- probabilistic framework
- probabilities
- probability
- probability density
- probability density function
- probability distribution
- procedure
- pronunciation
- recognition accuracy
- recognition error rate
- recognition errors
- reordering
- representations
- search space
- sentence
- sigmoid function
- sources of information
- speaker-independent continuous spelling task
- speech recognition errors
- stress
- technology
- temporal information
- term
- test set
- testing data
- text
- text corpus
- theories
- theory
- training
- training data
- training data set
- training set
- tree
- trigram
- trigram language model
- triphone
- true probability distribution
- utterance
- vocabulary
- word
- word error rate
- word error rates
- word model
- word sequence
- words
