ACL RD-TEC 1.0 Summarization of P04-1007
Paper Title:
DISCRIMINATIVE LANGUAGE MODELING WITH CONDITIONAL RANDOM FIELDS AND THE PERCEPTRON ALGORITHM
DISCRIMINATIVE LANGUAGE MODELING WITH CONDITIONAL RANDOM FIELDS AND THE PERCEPTRON ALGORITHM
Authors: Brian Roark and Murat Saraclar and Michael Collins and Mark Johnson
Primarily assigned technology terms:
- acoustic modeling
- algorithm
- approximation
- asr system
- automaton
- bayesian approach
- classifier
- communicator
- conditional likelihood
- conditional random field
- conditional random fields
- crf estimation
- crf training
- crfs
- darpa communicator
- decoding
- density estimation
- discriminative estimation
- discriminative language modeling
- discriminative training
- encoding
- estimation method
- feature selection
- finite state
- finite state automata
- finite-state automata
- hidden markov
- hidden markov model
- inner product
- language model modification
- language modeling
- large vocabulary speech recognition
- learning
- learning task
- likelihood estimation
- markov model
- maximum likelihood
- maximum likelihood estimation
- model estimation
- model modification
- modeling
- mutual information estimation
- nlp
- normalization
- optimization
- optimization method
- parameter estimation
- parameterization
- parsing
- perceptron
- perceptron algorithm
- perceptron training
- processor
- reading
- recognition
- recognizer
- regularization
- scoring
- search
- searching
- segmentation
- speech recognition
- speech recognizer
- support vector machines
- tagging
- training algorithm
- training method
- training process
- transcription
- transducers
- weighted automata
Other assigned terms:
- acoustic model
- acoustic models
- approach
- automata
- backoff
- baseline model
- bigram
- case
- conditional distribution
- conditional probabilities
- conditional probability
- convergence
- data set
- development set
- discriminative language model
- discriminative model
- distribution
- empirical results
- error rate
- estimation
- evaluation test
- evaluations
- experimental results
- fact
- feature
- feature set
- feature sets
- feature vector
- feature weight
- feature weights
- formalism
- gaussian prior
- generative model
- hypotheses
- hypothesis
- implementation
- index
- labeling
- language model
- language models
- large vocabulary speech
- lattice
- lattices
- lexicon
- likelihood
- log-likelihood
- mapping
- meaning
- measure
- method
- mutual information
- n-gram
- n-gram model
- n-gram models
- n-grams
- natural language
- natural language text
- nist
- nlp tasks
- norm
- optimization criterion
- oracle
- parameter values
- perplexity
- probabilities
- probability
- probability distribution
- procedure
- process
- recognition task
- semiring
- sentence
- sentences
- speech recognition task
- stress
- support vector
- symbol
- syntactic features
- system performance
- technique
- term
- test set
- text
- training
- training data
- training example
- training examples
- training set
- training text
- transcriptions
- transcripts
- trigram
- trigram model
- unbounded dependencies
- unigram
- utterance
- vocabulary
- word
- word error rate
- word sequences
- words
