ACL RD-TEC 1.0 Summarization of W03-1021
Paper Title:
TRAINING CONNECTIONIST MODELS FOR THE STRUCTURED LANGUAGE MODEL
TRAINING CONNECTIONIST MODELS FOR THE STRUCTURED LANGUAGE MODEL
Authors: Peng Xu and Ahmad Emami and Frederick Jelinek
Primarily assigned technology terms:
- algorithm
- approximation
- automatic speech recognition
- back-propagation
- back-propagation algorithm
- beam search
- binary branching
- clustering
- context clustering
- em algorithm
- em training
- encoding
- interpolated kneser-ney smoothing
- kneser-ney smoothing
- language modeling
- large vocabulary speech recognizer
- learning
- machine translation
- maximum entropy
- maximum entropy model
- modeling
- multi-stack search
- multi-stack search algorithm
- network training
- neural net
- neural network
- neural network training
- neural networks
- normalization
- parameter estimation
- parameterization
- parser
- parsing
- predictor
- probability estimation
- probability function
- recognition
- recognizer
- regularization
- search
- search algorithm
- searching
- smoothing
- smoothing method
- smoothing techniques
- speech recognition
- speech recognizer
- statistical machine translation
- stochastic parsing
- tagger
- training algorithm
- training procedure
- tuning
- word clustering
Other assigned terms:
- annotation
- approach
- baseline model
- beam
- bias
- case
- conditional probability
- conditional probability distribution
- connectionist
- corpora
- data sparseness
- data sparseness problem
- dimensionality
- distribution
- entropy
- error rate
- estimation
- events
- experimental results
- feature
- feature space
- feature vector
- feature vectors
- generation
- head word
- hypothesis
- hypothesis space
- implementation
- interpolation
- joint probability
- labeled training data
- language model
- language model performance
- language model probability
- language models
- large vocabulary speech
- learning rate
- lexical information
- likelihood
- likelihood function
- local maximum
- log-likelihood
- mapping
- method
- model parameters
- model performance
- model probability
- model size
- n-gram
- n-gram models
- natural speech
- parameter values
- parse
- parse structure
- part-of-speech
- part-of-speech set
- partial parse
- perplexity
- phrase
- pos tag
- preposition
- prepositional phrase
- probabilistic model
- probabilities
- probability
- probability distribution
- procedure
- representations
- search strategy
- sentence
- sentences
- sparseness problem
- structured language model
- syntactical information
- tags
- term
- terminals
- terms
- test data
- text
- toolkit
- training
- training corpus
- training criterion
- training data
- training examples
- tree
- treebank
- treebank annotation
- trees
- trigram
- trigram language model
- trigram model
- uniform distribution
- upenn treebank
- vocabulary
- word
- word error rate
- word sequence
- word string
- words
- wsj corpus
