1 / 20

Enriched translation model using morphology in MT

Enriched translation model using morphology in MT. Luong Minh Than g WING group meeting – 07 July, 2009. Overview. Brief recap on SMT & morphological analysis Motivation Enriched translation model Twin phrase-table construction Merging phrase tables Experiments Conclusion.

aspen
Download Presentation

Enriched translation model using morphology in MT

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Enriched translation model using morphology in MT Luong Minh Thang WING group meeting – 07 July, 2009

  2. Overview • Brief recap on SMT & morphological analysis • Motivation • Enriched translation model • Twin phrase-table construction • Merging phrase tables • Experiments • Conclusion

  3. SMT overview – alignment • Parallel data Ensinnäkin kohtaamiemme taloudellisten ja sosiaalisten vaikeuksien vuoksi on havaittavissa huolestumista , vaikka kasvu on kestävällä pohjalla ja tulosta vuosien ponnisteluista , kaikkien kansalaistemme taholta . These are , first and foremost , messages of concern at the economic and social problems that we are experiencing , in spite of a period of sustained growth stemming from years of efforts by all our fellow citizens . • Alignment: one-to-many (1-M) Source Target

  4. SMT overview – translation model • Intersect alignment 1-M + M-1  M – M • Extracting phrases from M-M alignment  translation model (phrase table). problems ||| ongelmat ||| 0.372611 0.597858 0.114146 0.13882 2.718 problems ||| ongelmasta ||| 0.352941 0.423077 0.000836237 0.0012435 2.718 … problems ||| vaikeuksista ||| 0.0696946 0.105991 0.0124042 0.0130002 2.718 problems ||| vaikeuksien ||| 0.0410959 0.062069 0.000836237 0.0010174 2.718 Translation probabilities Lexical probabilities Phrase penalty Foreign f English e

  5. Recap - Morphological analysis • Morpheme: minimal meaning-bearing unit English: machine + s, present + ed, etc. Finnish: oppositio + kansa + n + edusta + ja = opposition of parliament member • Morfessor (Creutz & Lagus, 2007): segment words, unsupervised manner un/PRE + fortunate/STM + ly/SUF

  6. Motivation • Problem: • Multiple word forms in morphology-complex language, e.g. ongelmat, ongelmasta, etc. • Rare words often occur and are hard to align  incorrect entries in normal (word-align) phrase table. • Solution: • Construct morpheme-align phrase table (PT) to aggregate better statistics for rare words. • Combine word- and morpheme-align PTs to produce even better translation model in a proper way.

  7. Overview • Brief recap on SMT & morphological analysis • Motivation • Enriched translation model • Twin phrase-table construction • Merging phrase tables • Experiments • Conclusion

  8. Twin phrase-table (PT) construction Word Morpheme GIZA++ GIZA++ Word alignment Morpheme alignment Phrase Extraction Phrase Extraction problem/STM+ s/SUF ||| ongelma/STM+ t/SUF problems ||| vaikeuksista PTm PTw Morphological segmentation PTwm problem/STM+ s/SUF ||| vaikeu/STM+ ksi/SUF+ sta/SUF PT merging Decoding

  9. Existing PT-merging methods • Add-feature - (Nakov, 2008; Chen et. al. 2009): F1 = F2 = F3 =  heuristic-driven • Interpolation - (Wu & Wang, 2007) : • tran(f|e) = α * tran1(f|e) + (1- α) * tran2(f|e) • lex(f|e) = β * lex1(f|e) + (1- β) * lex2(f|e)  not consider score “meaning” 1 if from 1st PT 1 if from 2nd PT 1 if from both PTs 0.5 otherwise 0.5 otherwise 0.5 otherwise

  10. Our merging method – normalizing translation probabilities problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| vaikeu + ksi + sta problem + s ||| vaikeu + ksi + sta problem + s ||| ongelma + sta PTwm PTm MLE tran1(e|f) =count1(e, f) / ∑e count1(e, f) tran2(e|f) =count2(e, f) / ∑e count2(e, f)

  11. Our merging method – normalizing translation probabilities problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| vaikeu + ksi + sta problem + s ||| vaikeu + ksi + sta problem + s ||| ongelma + sta PTwm PTm MLE tran(vaikeuksista | problems) =1/2=0.5 tran(ongelmasta | problems) =1/2=0.5 tran(ongelmat | problems) = 3/4 = 0.75 tran(vaikeuksista | problems) = 1/4 = 0.25 Interpolation (ratio = 0.5) tran(vaikeuksista | problems) = (0.5 + 0.25)/2 = 0.375 tran(ongelmat | problems) = (0 + 0.75)/2 = 0.375 tran(ongelmasta | problems) = (0.5 + 0)/2 = 0.25 Undesired translation!

  12. Our merging method – normalizing translation probabilities problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| vaikeu + ksi + sta problem + s ||| vaikeu + ksi + sta problem + s ||| ongelma + sta PTwm PTm MLE tran1(e|f) =count1(e, f) / ∑e count1(e, f) tran2(e|f) =count2(e, f) / ∑e count2(e, f) Normalization tran(e|f) =[ count1(e, f) + count2(e, f)] / [ ∑e count1(e, f) + ∑e count2(e, f) ]

  13. Our merging method – normalizing translation probabilities problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| ongelma + t problem + s ||| vaikeu + ksi + sta problem + s ||| vaikeu + ksi + sta problem + s ||| ongelma + sta PTwm PTm MLE tran(vaikeuksista | problems) =1/2=0.5 tran(ongelmasta | problems) =1/2=0.5 tran(ongelmat | problems) = 3/4 = 0.75 tran(vaikeuksista | problems) = 1/4 = 0.25 Normalization tran(vaikeuksista | problems) = (1 + 1)/(2+4) = 0.33 tran(ongelmat | problems) = (0 + 3)/(2 + 4) = 0.5 tran(ongelmasta | problems) = (1 + 0)/(2 + 4) = 0.17 Desired translation!

  14. Our merging method – full lexical probability interpolation lex(vaikeuksista | problems) = w1 lex(ongelmasta | problems) = w2 lex(vaikeu + ksi + sta | problem + s) = m1 lex(ongelma + t | problem + s) = m3 PTw lexical model PTm lexical model P(vaikeuksista|problems) P(ongelmasta|problems) P(vaikeu|problem), P(ongelma|problem), P(t|s), P(ksi|s),P(sta|s) Normal Interpolation (ratio = 0.5) Missing interpolated probabilities ! lex(vaikeuksista | problems) = (w1 + m1)/2 lex(ongelmat | problems) = (w2 + 0)/2 lex(ongelmasta | problems) = (0 + m3) /2 • Estimate lex(ongelma + sta | problem + s) using PTm lexical model  m2 • Estimate lex(ongelmat | problems) using PTw lexical model  w3 Full Interpolation

  15. Overview • Brief recap on SMT & morphological analysis • Motivation • Enriched translation model • Twin phrase-table construction • Merging phrase tables • Experiments • Conclusion

  16. Experiments – dataset • 2005 ACL shared task (Koehn & Monz, 2005)

  17. Experiments – baselines • w-system: uses PTw translate at word-level • m-system: uses PTm translate at morpheme-level • m-BLEU: BLEU where each token unit is a morpheme

  18. Experiments – our system • Improvements over m-system and w-system are statistically significant using sign test by (Collins et al. 2005)

  19. Conclusion Our contributions: • Enrich the translation model without using additional data. • Propose a principal way to merge phrase tables generated at different granularities.

  20. Q & A • Thank you !!!

More Related