ACL RD-TEC 1.0 Summarization of H93-1083
Paper Title:
SEGMENT-BASED ACOUSTIC MODELS FOR CONTINUOUS SPEECH RECOGNITION
SEGMENT-BASED ACOUSTIC MODELS FOR CONTINUOUS SPEECH RECOGNITION
Authors: Mari Ostendorf and J. Robin Rohlicek
Primarily assigned technology terms:
- acoustic modeling
- classification
- clustering
- context-dependent modeling
- continuous speech recognition
- likelihood ratio test
- maximum entropy
- mixture training
- modeling
- multipass search
- n-best rescoring
- ratio test
- recognition
- recognition system
- rescoring
- resource management
- score combination
- scoring
- search
- segmentation
- signal proceasing
- speaker-independent recognition
- speech recognition
- word recognition
Other assigned terms:
- acoustic models
- continuous speech
- correlation
- distribution
- entropy
- feature
- feature vectors
- formalism
- frame
- hierarchical structure
- hypotheses
- lattice
- likelihood
- likelihood ratio
- mapping
- phoneme
- resource management task
- rm corpus
- segments
- sentence
- signal
- technique
- test set
- timit corpus
- training
- utterance
- vocabulary
- word
