ACL RD-TEC 1.0 Summarization of H90-1067
Paper Title:
EXPERIMENTS WITH TREE-STRUCTURED MMI ENCODERS ON THE RM TASK
EXPERIMENTS WITH TREE-STRUCTURED MMI ENCODERS ON THE RM TASK
Authors: Mark T. Anikst and William S. Meisel and Matthew C. Soares and Kai-Fu Lee
Primarily assigned technology terms:
- binary decision tree
- classification
- context-dependent modeling
- continuous speech recognition
- data reduction
- decision tree
- decision trees
- decision-tree
- decoding
- dynamic programming
- encoder
- encoding
- forward-backward training
- hidden markov
- hidden markov model
- hidden markov models
- hmms
- markov model
- microprocessor
- modeling
- optimization
- processing
- pruning
- recognition
- recognition system
- recognition system \
- recognizer
- resource management
- search
- segmentation
- smoothing
- speaker-independent training
- speech recognition
- speech recognition system
- speech systems
- supervised training
- temporal resolution
- tree-growing
- vector quantization
- viterbi
- viterbi alignment
Other assigned terms:
- alphabet
- binary tree
- case
- classification tasks
- co-occurrence
- codebook
- context dependency
- continuous speech
- duration
- entropy
- error rate
- euclidean distance
- feature
- feature set
- feature space
- feature vector
- feature vectors
- frame
- frame tree
- grammar
- grammar \
- interpolation
- joint probability
- labeling
- leaf
- linear combination
- markov models
- measure
- mutual information
- optimization criterion
- perplexity
- probabilities
- probability
- procedure
- process
- recognition accuracy
- recognition errors
- segment boundaries
- segment duration
- segments
- sentences
- speech data
- subtree
- technology
- test set
- theory
- topology
- training
- training criterion
- training data
- training samples
- training set
- transcriptions
- tree
- tree frame
- trees
- triphone
- word
- word error rate
- word-pair grammar
