ACL RD-TEC 1.0 Summarization of N04-4021
Paper Title:
FEATURE-BASED PRONUNCIATION MODELING FOR SPEECH RECOGNITION
FEATURE-BASED PRONUNCIATION MODELING FOR SPEECH RECOGNITION
Authors: Karen Livescu and James Glass
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- automatic speech recognition
- decoding
- dynamic bayesian networks
- expectation-maximization
- hidden markov
- hidden markov models
- hmms
- learning
- modeling
- parameter learning
- pronunciation modeling
- recognition
- recognition improvement
- recognizer
- speech recognition
- synchronization
- transcription
- viterbi
- viterbi alignment
Other assigned terms:
- acoustic model
- approach
- case
- conversational speech
- conversational speech corpus
- data set
- development set
- dictionary
- distribution
- error rate
- experimental results
- fact
- feature
- feature set
- feature vectors
- feature-based pronunciation
- feature-based pronunciation model
- frame
- gesture
- index
- lattice
- lexical entry
- lexicon
- linguistic
- linguistic feature
- linguistic features
- mapping
- mappings
- markov models
- nasalization
- phoneme
- phonemes
- phonological rules
- probabilities
- probability
- procedure
- pronunciation
- representations
- rule set
- speech corpus
- switchboard corpus
- symbol
- terms
- test set
- text
- toolkit
- transcriptions
- transformation
- vocabulary
- vocal tract
- vowel
- word
- word error rate
- words
