Using sub-word n-gram models for dealing with OOV in large vocabulary speech recognition for Latvian.
Askars SalimbajevsJevgenijs StriginsPublished in: NODALIDA (2015)
Keyphrases
- n gram
- speech recognition
- language model
- out of vocabulary
- speech recognizer
- probabilistic model
- speech recognition systems
- word segmentation
- language modeling
- statistical language modeling
- language independent
- translation model
- acoustic models
- character n grams
- speech synthesis
- hidden markov models
- speech signal
- automatic speech recognition
- variable length
- document retrieval
- text classification
- retrieval model
- speech recognizers
- information retrieval
- pattern recognition
- relevance model
- speaker independent
- speaker identification
- spoken document retrieval
- machine learning
- speaker adaptation
- query expansion
- word level
- broadcast news
- test collection
- noisy environments
- language specific
- maximum likelihood
- image processing
- word recognition
- data mining
- part of speech