ACL RD-TEC 1.0 Summarization of H90-1060
Paper Title:
A NEW PARADIGM FOR SPEAKER-INDEPENDENT TRAINING AND SPEAKER ADAPTATION
A NEW PARADIGM FOR SPEAKER-INDEPENDENT TRAINING AND SPEAKER ADAPTATION
Authors: Francis Kubala and Richard Schwartz
Primarily assigned technology terms:
- 1000-word continuous speech recognition
- algorithm
- continuous speech recognition
- cooccurrence smoothing
- corrective training
- data collection
- database
- forward-backward algorithm
- hidden markov
- hidden markov models
- joint training
- large vocabulary continuous speech recognition
- modeling
- modeling technique
- processing
- re-estimation
- re-estimation procedure
- recognition
- resource management
- si recognition
- smoothing
- speaker adaptation
- speaker-independent training
- speaker-specific modeling
- speech recognition
- spoken language system
- spoken language systems
- training procedure
- transcription
- triphone cooccurrence smoothing
Other assigned terms:
- approach
- byblos system
- case
- cluster
- codebook
- context models
- continuous speech
- corpora
- distribution
- error rate
- experimental results
- feature
- feature space
- grammar
- interpolation
- lambda
- mapping
- markov models
- method
- model combination
- model parameters
- orthographic transcription
- perplexity
- phonemes
- probabilities
- procedure
- recognition task
- sentence
- sentences
- speech corpus
- speech data
- speech recognition task
- spoken language
- statistics
- technique
- test set
- trained model
- training
- training data
- transformation
- triphone
- utterance
- vocabulary
- word
- word error rate
- word-pair grammar
- words
