ACL RD-TEC 1.0 Summarization of W03-1019
Paper Title:
INVESTIGATING LOSS FUNCTIONS AND OPTIMIZATION METHODS FOR DISCRIMINATIVE LEARNING OF LABEL SEQUENCES
INVESTIGATING LOSS FUNCTIONS AND OPTIMIZATION METHODS FOR DISCRIMINATIVE LEARNING OF LABEL SEQUENCES
Authors: Yasemin Altun and Mark Johnson and Thomas Hofmann
Primarily assigned technology terms:
- accent prediction
- adaboost
- algorithm
- approximation
- binary classification
- boosting
- boosting algorithm
- chunking
- classification
- classifier
- classifiers
- conditional random fields
- crfs
- cross-validation
- discriminative approach
- discriminative learning
- discriminative modeling
- dual perceptron
- dynamic programming
- entity recognition
- exponential loss function
- feature mapping
- forward-backward algorithm
- function optimization
- hidden markov
- hidden markov model
- hmms
- iterative scaling
- kernel
- label sequence learning
- learning
- learning algorithm
- learning framework
- learning task
- loss function
- machine learning
- markov model
- maximum entropy
- maximum likelihood
- maximum likelihood optimization
- modeling
- named entity recognition
- named-entity recognition
- nlp
- normalization
- online perceptron
- optimization
- optimization method
- parameter estimation
- parameterization
- parsing
- perceptron
- perceptron algorithm
- pitch accent prediction
- pos tagging
- recognition
- recognizer
- regularization
- sequence boosting
- sequence labeling
- sequence learning
- sequential learning
- sequential optimization
- shallow parsing
- spelling
- supervised learning
- tagger
- tagging
- viterbi
Other assigned terms:
- accent
- approach
- bias
- bigram
- bigram model
- binary classification problem
- case
- chunks
- classification problem
- community
- computational complexity
- conditional independence
- conditional probability
- conditional probability distribution
- convergence
- corpora
- data sets
- dictionary
- distribution
- entropy
- entropy models
- error rate
- estimation
- evaluation function
- exp-loss function
- experimental results
- feature
- feature sets
- feature space
- feature weights
- gaussian prior
- generative model
- generative models
- independence assumption
- joint probability
- joint probability distribution
- labeling
- large corpora
- learning domain
- likelihood
- linear combination
- long distance dependencies
- mapping
- maps
- markov models
- maximum entropy models
- meaning
- method
- named entity
- named-entity
- names
- natural languages
- nlp community
- nlp tasks
- noise
- parallelism
- part of speech
- part-of-speech
- penn treebank
- penn treebank corpus
- pitch
- pitch accent
- probabilities
- probability
- probability distribution
- projection
- sentence
- sentences
- sequential exp-loss function
- speech tag
- statistic
- statistics
- svms
- tags
- term
- terms
- test data
- training
- training and test data
- training corpus
- training data
- training example
- training instance
- training set
- training time
- treebank
- treebank corpus
- window size
- word
- words
