ACL RD-TEC 1.0 Summarization of W06-1646
Paper Title:
CORRECTIVE MODELS FOR SPEECH RECOGNITION OF INFLECTED LANGUAGES
CORRECTIVE MODELS FOR SPEECH RECOGNITION OF INFLECTED LANGUAGES
Authors: Izhak Shafran and Keith Hall
Primarily assigned technology terms:
- acoustic model training
- algorithm
- analyzer
- asr system
- automatic speech recognition
- categorization
- computational linguistics
- computing
- corrective modeling
- czech speech recognition
- decoder
- decoding
- discriminative framework
- encoding
- entropy estimation
- entropy training
- error reduction
- estimation procedure
- extrinsic evaluation
- factoring
- feature extraction
- feature selection
- feature selection mechanism
- feature selection method
- finite state
- finite state machine
- genetic algorithm
- information technology
- language modeling
- language processing
- learning
- learning procedure
- linear regression
- maximum entropy
- maximum entropy framework
- maximum likelihood
- model size reduction
- model training
- modeling
- morphological analysis
- morphological analyzer
- morphology
- mutual information estimation
- natural language processing
- numerical optimization
- optimization
- parser
- perceptron
- perceptron algorithm
- processing
- re-estimation
- recognition
- recognition systems
- regression
- regularization
- reranking
- selection mechanism
- selection method
- speech recognition
- speech recognition systems
- tagger
- tagging
- text categorization
- training procedure
- transducer
- word bigram
Other assigned terms:
- acoustic model
- acoustic models
- adjective
- approach
- asr output
- association for computational linguistics
- backoff
- baseline model
- bias
- bigram
- case
- characters
- comparable corpora
- conditional model
- contextual features
- contingency table
- corpora
- czech morphology
- czech national corpus
- data sparsity
- dependency treebank
- development set
- discriminative language model
- distribution
- edit distance
- empirical evaluation
- empirical results
- entropy
- error rate
- estimation
- f-score
- factored language model
- feature
- feature sets
- feature space
- gaussian mixture
- hypotheses
- hypothesis
- implementation
- inflection
- inflectional morphology
- information gain
- input string
- interpretation
- knowledge
- language model
- language models
- large feature space
- lattice
- lattices
- lemma
- lexical items
- lexicon
- likelihood
- linguistics
- log-likelihood
- measures
- mechanisms
- method
- model parameters
- model size
- modifier
- morpheme
- morphemes
- morphological features
- morphological inflection
- mutual information
- n-gram
- n-gram language model
- n-grams
- natural language
- nist
- nouns
- number agreement
- oracle
- part-of-speech
- part-of-speech tags
- pragmatic interpretation
- prague dependency treebank
- procedure
- process
- pronunciation
- recognition accuracy
- recognition task
- representations
- semantic
- sentence
- speech recognition task
- statistics
- syntactic features
- tag model
- tags
- technology
- term
- test set
- text
- tokens
- training
- training set
- training time
- transcripts
- treebank
- trigram
- trigram language model
- triphone
- understanding
- unigram
- utterance
- vocabulary
- word
- word error rate
- word lattices
- word order
- word sequences
- wordform
- words
