AVENUE/LETRAS: Learning-based MT Approaches for Languages with Limited Resources

1. AVENUE/LETRAS:Learning-based MT Approaches for Languages with Limited Resources Alon Lavie, Jaime Carbonell, Lori Levin, Bob Frederking Joint work with: Erik Peterson, Christian Monson, Ariadna Font-Llitjos, Alison Alvarez, Roberto Aranovich

2. Why Machine Translation for Languages with Limited Resources? • We are in the age of information explosion • The internet+web+Google anyone can get the information they want anytime… • But what about the text in all those other languages? • How do they read all this English stuff? • How do we read all the stuff that they put online? • MT for these languages would Enable: • Better government access to native indigenous and minority communities • Better minority and native community participation in information-rich activities (health care, education, government) without giving up their languages. • Civilian and military applications (disaster relief) • Language preservation Learning-based MT with Limited Resources

3. AVENUE/LETRAS Funding • Started in 2000 with small amount of DARPA/TIDES funding (NICE) • AVENUE project funded by 5-year NSF ITR grant (2001-2006) • Follow-on LETRAS project funded by NSF HLC Program grant (2006-2009) • Collaboration funding sources: • Mapudungun (MINEDUC, Chile) • Hebrew (ISF, Israel) • Brazilian Portuguese & Native Langs. (Brazilian Gov.) • Inupiaq (NSF, Polar Programs) Learning-based MT with Limited Resources

4. CMU’s AVENUE Approach • Elicitation: use bilingual native informants to create a small high-quality word-aligned bilingual corpus of translated phrases and sentences • Building Elicitation corpora from feature structures • Feature Detection and Navigation • Transfer-rule Learning: apply ML-based methods to automatically acquire syntactic transfer rules for translation between the two languages • Learn from major language to minor language • Translate from minor language to major language • XFER + Decoder: • XFER engine produces a lattice of possible transferred structures at all levels • Decoder searches and selects the best scoring combination • Rule Refinement: refine the acquired rules via a process of interaction with bilingual informants • Morphology Learning • Word and Phrase bilingual lexicon acquisition Learning-based MT with Limited Resources

5. AVENUE MT Approach Semantic Analysis Sentence Planning Interlingua Syntactic Parsing Transfer Rules Text Generation AVENUE: Automate Rule Learning Source (e.g. Quechua) Target (e.g. English) Direct: SMT, EBMT Learning-based MT with Limited Resources

6. Learning Module Learning Module Learned Transfer Rules Handcrafted rules Morphology Analyzer AVENUE Architecture Elicitation Morphology Rule Learning Run-Time System RuleRefinement Translation Correction Tool Word-Aligned Parallel Corpus INPUT TEXT Run Time Transfer System Rule Refinement Module Elicitation Corpus Decoder Elicitation Tool Lexical Resources OUTPUT TEXT Learning-based MT with Limited Resources

7. Type information Part-of-speech/constituent information Alignments x-side constraints y-side constraints xy-constraints, e.g. ((Y1 AGR) = (X1 AGR)) Transfer Rule Formalism ;SL: the old man, TL: ha-ish ha-zaqen NP::NP [DET ADJ N] -> [DET N DET ADJ] ( (X1::Y1) (X1::Y3) (X2::Y4) (X3::Y2) ((X1 AGR) = *3-SING) ((X1 DEF = *DEF) ((X3 AGR) = *3-SING) ((X3 COUNT) = +) ((Y1 DEF) = *DEF) ((Y3 DEF) = *DEF) ((Y2 AGR) = *3-SING) ((Y2 GENDER) = (Y4 GENDER)) ) Learning-based MT with Limited Resources

8. Value constraints Agreement constraints Transfer Rule Formalism (II) ;SL: the old man, TL: ha-ish ha-zaqen NP::NP [DET ADJ N] -> [DET N DET ADJ] ( (X1::Y1) (X1::Y3) (X2::Y4) (X3::Y2) ((X1 AGR) = *3-SING) ((X1 DEF = *DEF) ((X3 AGR) = *3-SING) ((X3 COUNT) = +) ((Y1 DEF) = *DEF) ((Y3 DEF) = *DEF) ((Y2 AGR) = *3-SING) ((Y2 GENDER) = (Y4 GENDER)) ) Learning-based MT with Limited Resources

9. Transfer Rules  Transfer Trees NP PP NP1 NP P Adj N N1 ke eka aXyAya N jIvana NP NP1 PP Adj N P NP one chapter of N1 N life ; NP1 ke NP2 -> NP2 of NP1 ; Ex: jIvana ke eka aXyAya ; life of (one) chapter ; ==> a chapter of life ; {NP,12} NP::NP : [PP NP1] -> [NP1 PP] ( (X1::Y2) (X2::Y1) ; ((x2 lexwx) = 'kA') ) {NP,13} NP::NP : [NP1] -> [NP1] ( (X1::Y1) ) {PP,12} PP::PP : [NP Postp] -> [Prep NP] ( (X1::Y2) (X2::Y1) ) Learning-based MT with Limited Resources

10. Rule Learning - Overview • Goal: Acquire Syntactic Transfer Rules • Use available knowledge from the source side (grammatical structure) • Three steps: • Flat Seed Generation: first guesses at transfer rules; flat syntactic structure • Compositionality Learning:use previously learned rules to learn hierarchical structure • Constraint Learning: refine rules by learning appropriate feature constraints Learning-based MT with Limited Resources

11. Flat Seed Rule Generation Learning-based MT with Limited Resources

12. Compositionality Learning Learning-based MT with Limited Resources

13. Constraint Learning Learning-based MT with Limited Resources

14. AVENUE Prototypes • General XFER framework under development for past three years • Prototype systems so far: • German-to-English, Dutch-to-English • Chinese-to-English • Hindi-to-English • Hebrew-to-English • Portuguese-to-English • In progress or planned: • Mapudungun-to-Spanish • Quechua-to-Spanish • Inupiaq-to-English • Native-Brazilian languages to Brazilian Portuguese Learning-based MT with Limited Resources

15. Mapudungun • Indigenous Language of Chile and Argentina • ~ 1 Million Mapuche Speakers Learning-based MT with Limited Resources

16. Collaboration Eliseo Cañulef Rosendo Huisca Hugo Carrasco Hector Painequeo Flor Caniupil Luis Caniupil Huaiquiñir Marcela Collio Calfunao Cristian Carrillan Anton Salvador Cañulef • Mapuche Language Experts • Universidad de la Frontera (UFRO) • Instituto de Estudios Indígenas (IEI) • Institute for Indigenous Studies • Chilean Funding • Chilean Ministry of Education (Mineduc) • Bilingual and Multicultural Education Program Carolina Huenchullan Arrúe Claudio Millacura Salas Learning-based MT with Limited Resources

17. Accomplishments • Corpora Collection • Spoken Corpus • Collected: Luis Caniupil Huaiquiñir • Medical Domain • 3 of 4 Mapudungun Dialects • 120 hours of Nguluche • 30 hours of Lafkenche • 20 hours of Pwenche • Transcribed in Mapudungun • Translated into Spanish • Written Corpus • ~ 200,000 words • Bilingual Mapudungun – Spanish • Historical and newspaper text nmlch-nmjm1_x_0405_nmjm_00: M: <SPA>no pütokovilu kay ko C: no, si me lo tomaba con agua M: chumgechi pütokoki femuechi pütokon pu <Noise> C: como se debe tomar, me lo tomé pués nmlch-nmjm1_x_0406_nmlch_00: M: Chengewerkelafuymiürke C: Ya no estabas como gente entonces! Learning-based MT with Limited Resources

18. Accomplishments • Developed At UFRO • Bilingual Dictionary with Examples • 1,926 entries • Spelling Corrected Mapudungun Word List • 117,003 fully-inflected word forms • Segmented Word List • 15,120 forms • Stems translated into Spanish Learning-based MT with Limited Resources

19. Accomplishments • Developed at LTI using Mapudungun language resources from UFRO • Spelling Checker • Integrated into OpenOffice • Hand-built Morphological Analyzer • Prototype Machine Translation Systems • Rule-Based • Example-Based • Website: LenguasAmerindias.org Learning-based MT with Limited Resources

20. Challenges for Hebrew MT • Paucity in existing language resources for Hebrew • No publicly available broad coverage morphological analyzer • No publicly available bilingual lexicons or dictionaries • No POS-tagged corpus or parse tree-bank corpus for Hebrew • No large Hebrew/English parallel corpus • Scenario well suited for CMU transfer-based MT framework for languages with limited resources Learning-based MT with Limited Resources

21. Hebrew-to-English MT Prototype • Initial prototype developed within a two month intensive effort • Accomplished: • Adapted available morphological analyzer • Constructed a preliminary translation lexicon • Translated and aligned Elicitation Corpus • Learned XFER rules • Developed (small) manual XFER grammar as a point of comparison • System debugging and development • Evaluated performance on unseen test data using automatic evaluation metrics Learning-based MT with Limited Resources

22. Source Input בשורה הבאה Preprocessing Morphology Transfer Rules English Language Model {NP1,3} NP1::NP1 [NP1 "H" ADJ] -> [ADJ NP1] ((X3::Y1) (X1::Y2) ((X1 def) = +) ((X1 status) =c absolute) ((X1 num) = (X3 num)) ((X1 gen) = (X3 gen)) (X0 = X1)) Transfer Engine Translation Lexicon Decoder N::N |: ["$WR"] -> ["BULL"] ((X1::Y1) ((X0 NUM) = s) ((Y0 lex) = "BULL")) N::N |: ["$WRH"] -> ["LINE"] ((X1::Y1) ((X0 NUM) = s) ((Y0 lex) = "LINE")) Translation Output Lattice (0 1 "IN" @PREP) (1 1 "THE" @DET) (2 2 "LINE" @N) (1 2 "THE LINE" @NP) (0 2 "IN LINE" @PP) (0 4 "IN THE NEXT LINE" @PP) English Output in the next line

23. Morphology Example • Input word: B$WRH 0 1 2 3 4 |--------B$WRH--------| |-----B-----|$WR|--H--| |--B--|-H--|--$WRH---| Learning-based MT with Limited Resources

24. Morphology Example Y0: ((SPANSTART 0) Y1: ((SPANSTART 0) Y2: ((SPANSTART 1) (SPANEND 4) (SPANEND 2) (SPANEND 3) (LEX B$WRH) (LEX B) (LEX $WR) (POS N) (POS PREP)) (POS N) (GEN F) (GEN M) (NUM S) (NUM S) (STATUS ABSOLUTE)) (STATUS ABSOLUTE)) Y3: ((SPANSTART 3) Y4: ((SPANSTART 0) Y5: ((SPANSTART 1) (SPANEND 4) (SPANEND 1) (SPANEND 2) (LEX $LH) (LEX B) (LEX H) (POS POSS)) (POS PREP)) (POS DET)) Y6: ((SPANSTART 2) Y7: ((SPANSTART 0) (SPANEND 4) (SPANEND 4) (LEX $WRH) (LEX B$WRH) (POS N) (POS LEX)) (GEN F) (NUM S) (STATUS ABSOLUTE)) Learning-based MT with Limited Resources

25. Example Translation • Input: • לאחר דיונים רבים החליטה הממשלה לערוך משאל עם בנושא הנסיגה • After debates many decided the government to hold referendum in issue the withdrawal • Output: • AFTER MANY DEBATES THE GOVERNMENT DECIDED TO HOLD A REFERENDUM ON THE ISSUE OF THE WITHDRAWAL Learning-based MT with Limited Resources

26. Sample Output (dev-data) maxwell anurpung comes from ghana for israel four years ago and since worked in cleaning in hotels in eilat a few weeks ago announced if management club hotel that for him to leave israel according to the government instructions and immigration police in a letter in broken english which spread among the foreign workers thanks to them hotel for their hard work and announced that will purchase for hm flight tickets for their countries from their money Learning-based MT with Limited Resources

27. Challenges and Future Directions • Automatic Transfer Rule Learning: • Learning mappings for non-compositional structures • Effective models for rule scoring for • Decoding: using scores at runtime • Pruning the large collections of learned rules • Learning Unification Constraints • In the absence of morphology or POS annotated lexica • Integrated Xfer Engine and Decoder • Improved models for scoring tree-to-tree mappings, integration with LM and other knowledge sources in the course of the search Learning-based MT with Limited Resources

28. Challenges and Future Directions • Our approach for learning transfer rules is applicable to the large parallel data scenario, subject to solutions for several big challenges: • No elicitation corpus  break-down parallel sentences into reasonable learning examples • Working with less reliable automatic word alignments rather than manual alignments • Effective use of reliable parse structures for ONE language (i.e. English) and automatic word alignments in order to decompose the translation of a sentence into several compositional rules. • Effective scoring of resulting very large transfer grammars, and scaled up transfer + decoding Learning-based MT with Limited Resources

29. Future Research Directions • Automatic Rule Refinement • Morphology Learning • Feature Detection and Corpus Navigation • … Learning-based MT with Limited Resources

30. Implications for MT with Vast Amounts of Parallel Data • Phrase-to-phrase MT ill suited for long-range reorderings  ungrammatical output • Recent work on hierarchical Stat-MT [Chiang, 2005] and parsing-based MT [Melamed et al, 2005] [Knight et al] • Learning general tree-to-tree syntactic mappings is equally problematic: • Meaning is a hybrid of complex, non-compositional phrases embedded within a syntactic structure • Some constituents can be translated in isolation, others require contextual mappings Learning-based MT with Limited Resources

31. Test set of 62 sentences from Haaretz newspaper, 2 reference translations Evaluation Results Learning-based MT with Limited Resources

32. Hebrew-English: Test Suite Evaluation Learning-based MT with Limited Resources

33. QuechuaSpanish MT • V-Unit: funded Summer project in Cusco (Peru) June-August 2005 [preparations and data collection started earlier] • Intensive Quechua course in Centro Bartolome de las Casas (CBC) • Worked together with two Quechua native and one non-native speakers on developing infrastructure (correcting elicited translations, segmenting and translating list of most frequent words) Learning-based MT with Limited Resources

34. Quechua  Spanish Prototype MT System • Stem Lexicon (semi-automatically generated): 753 lexical entries • Suffix lexicon: 21 suffixes • (150 Cusihuaman) • Quechua morphology analyzer • 25 translation rules • Spanish morphology generation module • User-Studies: 10 sentences, 3 users (2 native, 1 non-native) Learning-based MT with Limited Resources

35. The Transfer Engine Learning-based MT with Limited Resources

36. The Transfer Engine • Some Unique Features: • Works with either learned or manually-developed transfer grammars • Handles rules with or without unification constraints • Supports interfacing with servers for Morphological analysis and generation • Can handle ambiguous source-word analyses and/or SL segmentations represented in the form of lattice structures Learning-based MT with Limited Resources

37. The Lattice Decoder • Simple Stack Decoder, similar in principle to SMT/EBMT decoders • Searches for best-scoring path of non-overlapping lattice arcs • Scoring based on log-linear combination of scoring components (no MER training yet) • Scoring components: • Standard trigram LM • Fragmentation: how many arcs to cover the entire translation? • Length Penalty • Rule Scores (not fully integrated yet) Learning-based MT with Limited Resources

38. Outline • Rationale for learning-based MT • Roadmap for learning-based MT • Framework overview • Elicitation • Learning transfer Rules • Automatic rule refinement • Example prototypes • Implications for MT with vast parallel data • Conclusions and future directions Learning-based MT with Limited Resources

39. Data Elicitation for Languages with Limited Resources • Rationale: • Large volumes of parallel text not available  create a small maximally-diverse parallel corpus that directly supports the learning task • Bilingual native informant(s) can translate and align a small pre-designed elicitation corpus, using elicitation tool • Elicitation corpus designed to be typologically and structurally comprehensive and compositional • Transfer-rule engine and new learning approach support acquisition of generalized transfer-rules from the data Learning-based MT with Limited Resources

40. Elicitation Tool:English-Chinese Example Learning-based MT with Limited Resources

41. Elicitation Tool:English-Chinese Example Learning-based MT with Limited Resources

42. Elicitation Tool:English-Hindi Example Learning-based MT with Limited Resources

43. Elicitation Tool:English-Arabic Example Learning-based MT with Limited Resources

44. Elicitation Tool:Spanish-Mapudungun Example Learning-based MT with Limited Resources

45. Designing Elicitation Corpora • What do we want to elicit? • Diversity of linguistic phenomena and constructions • Syntactic structural diversity • How do we construct an elicitation corpus? • Typological Elicitation Corpus based on elicitation and documentation work of field linguists (e.g. Comrie 1977, Bouquiaux 1992): initial corpus size ~1000 examples • Structural Elicitation Corpus based on representative sample of English phrase structures: ~120 examples • Organized compositionally: elicit simple structures first, then use them as building blocks • Goal: minimize size, maximize linguistic coverage Learning-based MT with Limited Resources

46. Typological Elicitation Corpus • Feature Detection • Discover what features exist in the language and where/how they are marked • Example: does the language mark gender of nouns? How and where are these marked? • Method: compare translations of minimal pairs –sentences that differ in only ONE feature • Elicit translations/alignments for detected features and their combinations • Dynamic corpus navigation based on feature detection: no need to elicit for combinations involving non-existent features Learning-based MT with Limited Resources

47. Typological Elicitation Corpus • Initial typological corpus of about 1000 sentences was manually constructed • New construction methodology for building an elicitation corpus using: • A feature specification: lists inventory of available features and their values • A definition of the set of desired feature structures • Schemas define sets of desired combinations of features and values • Multiplier algorithm generates the comprehensive set of feature structures • A generation grammar and lexicon: NLG generator generates NL sentences from the feature structures Learning-based MT with Limited Resources

48. Structural Elicitation Corpus • Goal: create a compact diverse sample corpus of syntactic phrase structures in English in order to elicit how these map into the elicited language • Methodology: • Extracted all CFG “rules” from Brown section of Penn TreeBank (122K sentences) • Simplified POS tag set • Constructed frequency histogram of extracted rules • Pulled out simplest phrases for most frequent rules for NPs, PPs, ADJPs, ADVPs, SBARs and Sentences • Some manual inspection and refinement • Resulting corpus of about 120 phrases/sentences representing common structures • See [Probst and Lavie, 2004] Learning-based MT with Limited Resources

49. Outline • Rationale for learning-based MT • Roadmap for learning-based MT • Framework overview • Elicitation • Learning transfer Rules • Automatic rule refinement • Example prototypes • Implications for MT with vast parallel data • Conclusions and future directions Learning-based MT with Limited Resources

50. Flat Seed Rule Generation • Create a “flat” transfer rule specific to the sentence pair, partially abstracted to POS • Words that are aligned word-to-word and have the same POS in both languages are generalized to their POS • Words that have complex alignments (or not the same POS) remain lexicalized • One seed rule for each translation example • No feature constraints associated with seed rules (but mark the example(s) from which it was learned) Learning-based MT with Limited Resources