ACL RD-TEC 1.0 Summarization of P06-2101
Paper Title:
MINIMUM RISK ANNEALING FOR TRAINING LOG-LINEAR MODELS
MINIMUM RISK ANNEALING FOR TRAINING LOG-LINEAR MODELS
Authors: David A. Smith and Jason Eisner
Primarily assigned technology terms:
- alignment template model
- approximation
- bayes risk decoding
- classifier
- computational linguistics
- computing
- convex optimization
- decoder
- decoding
- dependency parsing
- dynamic programming
- error minimization
- giza
- labeled dependency parsing
- lagrange multiplier
- language modeling
- learner
- learning
- likelihood maximization
- line-search method
- linear programming
- loss function
- machine translation
- machine translation systems
- maximum likelihood
- measuring
- modeling
- multidimensional optimization
- multilingual dependency parsing
- natural language system
- nlp
- optimization
- parser
- parsers
- parsing
- part-of-speech tagging
- phrase extraction
- phrase-based decoder
- phrase-based machine translation
- posteriori estimation
- programming method
- recognition
- recognition system
- regularization
- rescoring
- risk minimization
- search
- smoothing
- speech recognition
- speech recognition system
- structured prediction
- supervised training
- tagging
- translation systems
- unsupervised learning
- viterbi
Other assigned terms:
- alignment template
- approach
- association for computational linguistics
- bleu
- bleu metric
- bleu score
- bleu scores
- brevity penalty
- community
- conditional probability
- conll-x
- convergence
- corpora
- development set
- distribution
- dutch
- entropy
- error rate
- estimation
- evaluation metrics
- evaluation set
- f-measure
- feature
- feature space
- feature vector
- french
- gaussian prior
- generative models
- geometric mean
- gold standard
- hypotheses
- hypothesis
- hypothesis lattice
- labeling
- language pair
- language pairs
- large corpus
- lattice
- likelihood
- linear combination
- linguistics
- local maxima
- log-likelihood
- log-linear combination
- log-linear model
- log-linear models
- meaning
- measures
- method
- model parameters
- n-gram
- natural language
- opinion
- parameter space
- parse
- part-of-speech
- parts of speech
- permutation
- phrase
- precision
- probabilities
- probability
- probability distribution
- probability estimates
- procedure
- process
- random sample
- reference translation
- sentence
- sentences
- statistics
- technique
- term
- test corpora
- test corpus
- test data
- test set
- theorem
- tokens
- training
- training corpus
- training criterion
- training data
- training examples
- training set
- transcripts
- translations
- trees
- uniform distribution
- unigram
- word
- word count
- word error rate
- words
