Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC...

27
Discriminative Modeling Discriminative Modeling extraction Sets for extraction Sets for Machine Translation Machine Translation Author John DeNero and Dan Klein UC Berkeley Presenter Justin Chiu

description

Progress of Statistical MT Generate translated sentences word by word Using while fragments of training example, building translation rules ◦ Aligned at the word level ◦ Extract fragment-level rules from word aligned sentence pair  Tree to string translation Extraction Set Models ◦ Set of all overlapping phrasal translation rule + alignment

Transcript of Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC...

Page 1: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Discriminative Modeling Discriminative Modeling extraction Sets for Machine extraction Sets for Machine TranslationTranslationAuthorJohn DeNero and Dan Klein UC BerkeleyPresenterJustin Chiu

Page 2: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

ContributionContributionExtraction set

◦Nested collections of all the overlapping phrase pairs consistent with an underlying word-alignment

Advantages over word-factored alignment model◦Can incorporate features on phrase pairs,

more than word link◦Optimize a extraction-based loss function

really direct to generating translationPerform better than both supervised

and unsupervised baseline

Page 3: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Progress of Statistical MTProgress of Statistical MTGenerate translated sentences

word by wordUsing while fragments of training

example, building translation rules◦Aligned at the word level ◦Extract fragment-level rules from word

aligned sentence pair Tree to string translation

Extraction Set Models◦Set of all overlapping phrasal

translation rule + alignment

Page 4: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

OutlineOutlineExtraction Set ModelsModel EstimationModel InferenceExperiments

Page 5: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

EXTRACTION SET EXTRACTION SET MODELSMODELS

Page 6: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Extraction Set ModelsExtraction Set ModelsInput

◦Unaligned sentence

Output◦Extraction set of phrasal translation

rules◦Word alignment

Page 7: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Extraction Sets from Word Extraction Sets from Word AlignmentsAlignments

Page 8: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Extraction Sets from Word Extraction Sets from Word AlignmentsAlignments

Page 9: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Extraction Sets from Word Extraction Sets from Word AlignmentsAlignments

Page 10: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Possible and Null Alignment Possible and Null Alignment LinksLinksPossible links has two types

◦ Function words that is unique in its language◦ Short phrase that has no lexical equivalent

Null alignment◦ Express content that is

absent in its translation

Page 11: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Interpreting Possible and Null Interpreting Possible and Null Alignment LinksAlignment Links

Page 12: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Interpreting Possible and Null Interpreting Possible and Null Alignment LinksAlignment Links

Page 13: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Linear Model for Linear Model for Extraction SetExtraction Set

Page 14: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Scoring Extraction SetsScoring Extraction Sets

Page 15: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

MODEL ESTIMATIONMODEL ESTIMATION

Page 16: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

MIRA(Margin-infused Relaxed MIRA(Margin-infused Relaxed Algorithm)Algorithm)

Page 17: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Extraction Set Loss Extraction Set Loss FunctionFunction

Page 18: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

MODEL INFERENCEMODEL INFERENCE

Page 19: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Possible DecompositionsPossible Decompositions

Page 20: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

DP for Extraction SetsDP for Extraction Sets

Page 21: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

DP for Extraction SetsDP for Extraction Sets

Page 22: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Finding Pseudo-Gold ITG Finding Pseudo-Gold ITG AlignmentAlignment

Page 23: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

EXPERIMENTSEXPERIMENTS

Page 24: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

Five systems for Five systems for comparisoncomparisonUnsupervised baseline◦ Giza++◦ Joint HMMSupervised baseline◦ Block ITGExtraction Set Coarse Pass◦ Does not score bispans that corss

bracketing of ITG derivationsFull Extraction Set Model

Page 25: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

DataDataDiscriminative training and

alignment evaluation◦Trained baseline HMM on 11.3 million

words of FBIS newswire data◦Hand-aligned portion of the NIST MT02

test set 150 training and 191 test sentences

End-to-end translation experiments◦Trained on 22.1 million word prarllel

corpus consisting of sentence up to 40 of newswire data from GALE program

◦NIST MT04/MT05 test sets

Page 26: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

ResultsResults

Page 27: Discriminative Modeling extraction Sets for Machine Translation Author John DeNero and Dan KleinUC Berkeley Presenter Justin Chiu.

DiscussionDiscussionSyntax labels v.s wordsWord align to rule Rule to word

alignInformation from two directions65% of type 1 error