Recognition of target domain Japanese speech using language model replacement.
Daiki MoriKengo OhtaRyota NishimuraAtsunori OgawaNorihide KitaokaPublished in: EURASIP J. Audio Speech Music. Process. (2024)
Keyphrases
- language model
- speech recognition
- word error rate
- domain adaptation
- target domain
- language modeling
- handwriting recognition
- transfer learning
- cross domain
- n gram
- document retrieval
- retrieval model
- probabilistic model
- query expansion
- labeled data
- test collection
- automatic speech recognition
- information retrieval
- mixture model
- training data
- smoothing methods
- pattern recognition
- speech signal
- unlabeled data
- error rate
- semi supervised learning
- feature extraction
- multiple sources
- text categorization
- test data
- context sensitive
- training samples
- learning algorithm
- information retrieval systems
- natural language
- hidden markov models
- knowledge discovery
- domain specific
- maximum likelihood
- data sets