ACL RD-TEC 1.0 Summarization of W04-3237
Paper Title:
ADAPTATION OF MAXIMUM ENTROPY CAPITALIZER: LITTLE DATA CAN HELP A LO
ADAPTATION OF MAXIMUM ENTROPY CAPITALIZER: LITTLE DATA CAN HELP A LO
Authors: Ciprian Chelba and Alex Acero
Primarily assigned technology terms:
- algorithm
- automatic capitalization
- capitalization
- classification
- cut-off feature selection
- dynamic programming
- feature selection
- feature selection algorithm
- hmms
- information extraction
- language modeling
- language modeling approach
- machine translation
- map adaptation
- markov model
- maxent
- maximum entropy
- maximum entropy model
- maximum likelihood
- memm tagger
- modeling
- parsing
- part-of-speech tagging
- punctuation generation and capitalization
- recognition
- reestimation
- rule-based tagger
- search
- segmenter
- selection algorithm
- sequence labeling
- sequence tagging
- smoothing
- speech recognition
- tagger
- tagging
- text routing
- training procedure
Other assigned terms:
- approach
- background model
- broadcast news
- broadcast news data
- case
- classification accuracy
- conditional probability
- conditional probability model
- data sets
- derivation
- distribution
- entropy
- entropy markov model
- entropy models
- entropy probability model
- error rate
- exponential model
- fact
- feature
- feature set
- feature sets
- feature weights
- gaussian prior
- generation
- kl divergence
- labeling
- language model
- likelihood
- log-likelihood
- markov models
- maxent model
- maximum entropy models
- model parameters
- model size
- natural language
- parameter values
- part-of-speech
- prior distribution
- probabilistic approach
- probabilistic model
- probability
- probability model
- probability value
- procedure
- punctuation
- recognition errors
- rule-based model
- sentence
- sentences
- speech recognition errors
- state transition model
- statistics
- syntactic context
- tag sequence
- tagging accuracy
- tagging problem
- tags
- technique
- test data
- test set
- text
- training
- training data
- training phase
- training set
- vocabulary
- wall street journal text
- word
- word sequence
- word sequences
- words
