1 / 32

*Introduction to Natural Language Processing (600.465) Statistical Machine Translation

*Introduction to Natural Language Processing (600.465) Statistical Machine Translation. Dr. Jan Hajič cCS Dept., Johns Hopkins Univ. hajic@cs.jhu.edu www.cs.jhu.edu/~hajic. The Main Idea. Treat translation as a noisy channel problem:

bleigh
Download Presentation

*Introduction to Natural Language Processing (600.465) Statistical Machine Translation

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. *Introduction to Natural Language Processing (600.465)Statistical Machine Translation Dr. Jan Hajič cCS Dept., Johns Hopkins Univ. hajic@cs.jhu.edu www.cs.jhu.edu/~hajic JHU CS 600.465/Jan Hajic

  2. The Main Idea • Treat translation as a noisy channel problem: Input (Source) “Noisy” Output (target) The channel E: English words... (adds “noise”) F: Les mots Anglais... • TheModel: P(E|F) = P(F|E) P(E) / P(F) • Interested in rediscovering E given F: After the usual simplification (P(F) fixed): argmaxE P(E|F) = argmaxE P(F|E) P(E) ! JHU CS 600.465/ Intro to NLP/Jan Hajic

  3. The Necessities • Language Model (LM) P(E) • Translation Model (TM): Target given source P(F|E) • Search procedure • Given E, find best F using the LM and TM distributions. • Usual problem: sparse data • We cannot create a “sentence dictionary” E ↔F • Typically, we do not see a sentence even twice! JHU CS 600.465/ Intro to NLP/Jan Hajic

  4. The Language Model • Any LM will do: • 3-gram LM • 3-gram class-based LM • decision tree LM with hierarchical classes • Does not necessarily operates on word forms: • cf. later the “analysis” and “generation” procedures • for simplicity, imagine now it does operate on word forms JHU CS 600.465/ Intro to NLP/Jan Hajic

  5. The Translation Models • Do not care about correct strings of English words (that’s the task of the LM) • Therefore, we can make more independence assumptions: • for start, use the “tagging” approach: • 1 English word (“tag”) ~ 1 French word (“word”) • not realistic: rarely even the number of words is the same in both sentences (let alone there is 1:1 correspondence!) • Þ use “Alignment”. JHU CS 600.465/ Intro to NLP/Jan Hajic

  6. The Alignment 0 1 2 3 4 5 6 • e0 And the program has been implemented • f0 Le programme a été mis en application 0 1 2 3 4 5 6 7 • Linear notation: • f0(1) Le(2) programme(3) a(4) été(5) mis(6) en(6) application(6) • e0 And(0) the(1) program(2) has(3) been(4) implemented(5,6,7) JHU CS 600.465/ Intro to NLP/Jan Hajic

  7. Alignment Mapping • In general: • |F| = m, |E| = l (length of sent.): • lm connections (each French word to any English word), • 2lm different alignments for any pair (E,F) (any subset) • In practice: • From English to French • each English word 1-n connections (n - empirical max.-fertility?) • each French word exactly 1 connection • therefore, “only”(l+1)m alignments ( << 2lm ) • aj = i (link from j-th French word goes to i-th English word) JHU CS 600.465/ Intro to NLP/Jan Hajic

  8. Elements of Translation Model(s) • Basic distribution: • P(F,A,E) - the joint distribution of the English sentence, the Alignment, and the French sentence (length m) • Interested also in marginal distributions: P(F,E) = SA P(F,A,E) P(F|E) = P(F,E) / P(E) = SA P(F,A,E) / SA,F P(F,A,E) = SA P(F,A|E) • Useful decomposition [one of possible decompositions]: P(F,A|E) = P(m | E) Pj=1..m P(aj|a1j-1,f1j-1,m,E) P(fj|a1j,f1j-1,m,E) JHU CS 600.465/ Intro to NLP/Jan Hajic

  9. Decomposition • Decomposition formula again: P(F,A|E) = P(m | E) Pj=1..m P(aj|a1j-1,f1j-1,m,E) P(fj|a1j,f1j-1,m,E) m - length of French sentence aj - the alignment (single connection) going from j-th French w. fj - the j-th French word from F a1j-1 - sequence of alignments ai up to the word preceding fj a1j - sequence of alignments ai up to and including the word fj f1j-1 - sequence of French words up to the word preceding fj JHU CS 600.465/ Intro to NLP/Jan Hajic

  10. Decomposition and the Generative Model • ...and again: P(F,A|E) = P(m | E) Pj=1..m P(aj|a1j-1,f1j-1,m,E) P(fj|a1j,f1j-1,m,E) • Generate: • first, the length of the French given the English words E; • then, the link from the first position in F (not knowing the actual word yet) Þ now we know the English word • then, given the link (and thus the English word), generate the French word at the current position • then, move to the next position in F until m position filled. JHU CS 600.465/ Intro to NLP/Jan Hajic

  11. Approximations • Still too many parameters • similar situation as in n-gram model with “unlimited” n • impossible to estimate reliably. • Use 5 models, from the simplest to the most complex (i.e. from heavy independence assumptions to light) • Parameter estimation: Estimate parameters of Model 1; use as an initial estimate for estimating Model 2 parameters; etc. JHU CS 600.465/ Intro to NLP/Jan Hajic

  12. Model 1 • Approximations: • French length P(m | E) is constant (small e) • Alignment link distribution P(aj|a1j-1,f1j-1,m,E) depends on English length l only (= 1/(l+1)) • French word distribution depends only on the English and French word connected with link aj. • Þ Model 1 distribution: P(F,A|E) = e / (l+1)mPj=1..m p(fj|eaj) JHU CS 600.465/ Intro to NLP/Jan Hajic

  13. Models 2-5 • Model 2 • adds more detail into P(aj|...): more “vertical” links preferred • Model 3 • adds “fertility” (number of links for a given English word is explicitly modeled: P(n|ei) • “distortion” replaces alignment probabilities from Model 2 • Model 4 • the notion of “distortion” extended to chunks of words • Model 5 is Model 4, but not deficient (does not waste probability to non-strings) JHU CS 600.465/ Intro to NLP/Jan Hajic

  14. The Search Procedure • “Decoder”: • given “output” (French), discover “input” (English) • Translation model goes in the opposite direction: p(f|e) = .... • Naive methods do not work. • Possible solution (roughly): • generate English words one-by-one, keep only n-best (variable n) list; also, account for different lengths of the English sentence candidates! JHU CS 600.465/ Intro to NLP/Jan Hajic

  15. Analysis - Translation - Generation (ATG) • Word forms: too sparse • Use four basic analysis, generation steps: • tagging • lemmatization • word-sense disambiguation • noun-phrase “chunks” (non-compositional translations) • Translation proper: • use chunks as “words” JHU CS 600.465/ Intro to NLP/Jan Hajic

  16. Training vs. Test with ATG • Training: • analyze both languages using all four analysis steps • train TM(s) on the result (i.e. on chunks, tags, etc.) • train LM on analyzed source (English) • Runtime/Test: • analyze given language sentence (French) using identical tools as in training • translate using the trained Translation/Language model(s) • generate source (English), reversing the analysis process JHU CS 600.465/ Intro to NLP/Jan Hajic

  17. Analysis: Tagging and Morphology • Replace word forms by morphologically processed text: • lemmas • tags • original approach: mix them into the text, call them “words” • e.g. She bought two books. Þ she buy VBP two book NNS. • Tagging: yes • but reversed order: • tag first, then lemmatize [NB: does not work for inflective languages] • technically easy • Hand-written deterministic rules for tag+form Þ lemma JHU CS 600.465/ Intro to NLP/Jan Hajic

  18. Word Sense Disambiguation, Word Chunking • Sets of senses for each E, F word: • e.g. book-1, book-2, ..., book-n • prepositions (de-1, de-2, de-3,...), many others • Senses derived automatically using the TM • translation probabilities measured on senses: p(de-3|from-5) • Result: • statistical model for assigning senses monolingually based on context (also MaxEnt model used here for each word) • Chunks: group words for non-compositional translation JHU CS 600.465/ Intro to NLP/Jan Hajic

  19. Generation • Inverse of analysis • Much simpler: • Chunks Þ words (lemmas) with senses (trivial) • Words (lemmas) with senses Þ words (lemmas) (trivial) • Words (lemmas) + tags Þ word forms • Additional step: • Source-language ambiguity: • electric vs. electrical, hath vs. has, you vs. thou: treated as a single unit in translation proper, but must be disambiguated at the end of generation phase; using additional pure LM on word forms. JHU CS 600.465/ Intro to NLP/Jan Hajic

  20. *Introduction to Natural Language Processing (600.465)Statistical Translation: Alignment and Parameter Estimation Dr. Jan Hajič CS Dept., Johns Hopkins Univ. hajic@cs.jhu.edu www.cs.jhu.edu/~hajic JHU CS 600.465/Jan Hajic

  21. Alignment • Available corpus assumed: • parallel text (translation E ↔F) • No alignment present (day marks only)! • Sentence alignment • sentence detection • sentence alignment • Word alignment • tokenization • word alignment (with restrictions) JHU CS 600.465/ Intro to NLP/Jan Hajic

  22. Sentence Boundary Detection • Rules, lists: • Sentence breaks: • paragraphs (if marked) • certain characters: ?, !, ; (...almost sure) • The Problem: period . • could be end of sentence (... left yesterday. He was heading to...) • decimal point: 3.6 (three-point-six) • thousand segment separator: 3.200 (three-thousand-two-hundred) • abbreviation never at the end of sentence: cf., e.g., Calif., Mt., Mr. • ellipsis: ... • other languages: ordinal number indication (2nd ~ 2.) • initials: A. B. Smith • Statistical methods: e.g., Maximum Entropy JHU CS 600.465/ Intro to NLP/Jan Hajic

  23. Sentence Alignment • The Problem: sentences detected only: • E: • F: • Desired output: Segmentation with equal number of segments, spanning continuously the whole text. • Original sentence boundaries kept: • E: • F: • Alignments obtained: 2-1, 1-1, 1-1, 2-2, 2-1, 0-1 • New segments called “sentences” from now on. JHU CS 600.465/ Intro to NLP/Jan Hajic

  24. Alignment Methods • Several methods (probabilistic and not prob.) • character-length based • word-length based • “cognates” (word identity used) • using an existing dictionary (F: prendre ~ E: make, take) • using word “distance” (similarity): names, numbers, borrowed words, Latin origin words, ... • Best performing: • statistical, word- or character- length based (with some words perhaps) JHU CS 600.465/ Intro to NLP/Jan Hajic

  25. Length-based Alignment • First, define the problem probabilistically: argmaxA P(A|E,F) = argmaxA P(A,E,F) (E,F fixed) • Define a “bead”: • E: • F: • Approximate: P(A,E,F) @Pi=1..nP(Bi), where Bi is a bead; P(Bi) does not depend on the rest of E,F. “bead” (2:2 in this case) JHU CS 600.465/ Intro to NLP/Jan Hajic

  26. The Alignment Task • Given the model definition, P(A,E,F) @Pi=1..nP(Bi), find the partitioning of (E,F) into n beads Bi=1..n, that maximizes P(A,E,F) over training data. • Define Bi = p:qai, where p:q {0:1,1:0,1:1,1:2,2:1,2:2} • describes the type of alignment • Want to use some sort of dynamic programming: • Define Pref(i,j)... probability of the best alignment from the start of (E,F) data (1,1) up to (i,j) JHU CS 600.465/ Intro to NLP/Jan Hajic

  27. P(1:0ak) Pref(i-2,j-2) Pref(i-2,j-1) Pref(i-1,j-2) Pref(i-1,j-1) Pref(i-1,j) Pref(i,j-1) P(2:2ak) P(2:1ak) P(1:2ak) P(0:1ak) P(1:1ak) Recursive Definition • Initialize: Pref(0,0) = 0. • Pref(i,j) = max ( Pref(i,j-1) P(0:1ak), Pref(i-1,j) P(1:0ak), Pref(i-1,j-1) P(1:1ak), Pref(i-1,j-2) P(1:2ak), Pref(i-2,j-1) P(2:1ak), Pref(i-2,j-2) P(2:2ak) ) • This is enough for a Viterbi-like search. • E: • F: i j JHU CS 600.465/ Intro to NLP/Jan Hajic

  28. Probability of a Bead • Remains to define P(p:qak) (the red part): • k refers to the “next” bead, with segments of p and q sentences, lengths lk,e and lk,f. • Use normal distribution for length variation: • P(p:qak) = P(d(lk,e,lk,f,m,s2),p:q) @ P(d(lk,e,lk,f,m,s2))P(p:q) • d(lk,e,lk,f,m,s2) = (lk,f - mlk,e)/lk,es2 • Estimate P(p:q) from small amount of data, or even guess and re-estimate after aligning some data. • Words etc. might be used as better clues in P(p:qak) def. JHU CS 600.465/ Intro to NLP/Jan Hajic

  29. Saving time • For long texts (> 104 sentences), even Viterbi (in the version needed) is not effective (o(S2) time) • Go paragraph by paragraph if they are aligned 1:1 • What if not? • Apply the same method first to paragraphs! • identify paragraphs roughly in both languages • run the algorithm to get aligned paragraph-like segments • then, run on sentences within paragraphs. • Performs well if not many consecutive 1:0 or 0:1 beads. JHU CS 600.465/ Intro to NLP/Jan Hajic

  30. Word alignment • Length alone does not help anymore. • mainly because words can be swapped, and mutual translations have often vastly different length. • ...but at least, we have “sentences” (sentence-like segments) aligned; that will be exploited heavily. • Idea: • Assume some (simple) translation model (such as Model 1). • Find its parameters by considering virtually all alignments. • After we have the parameters, find the best alignment given those parameters. JHU CS 600.465/ Intro to NLP/Jan Hajic

  31. Word Alignment Algorithm • Start with sentence-aligned corpus. • Let (E,F) be a pair of sentences (actually, a bead). • Initialize p(f|e) randomly (e.g., uniformly), fF, eE. • Compute expected counts over the corpus: c(f,e) = S(E,F);eE,fF p(f|e) " aligned pair (E,F), find if e in E and f in F; if yes, add p(f|e). • Reestimate: p(f|e) = c(f,e) / c(e) [c(e) = Sf c(f,e)] • Iterate until change of p(f|e) is small. JHU CS 600.465/ Intro to NLP/Jan Hajic

  32. Best Alignment • Select, for each (E,F), A = argmaxA P(A|F,E) = argmaxAP(F,A|E)/P(F) = argmaxA P(F,A|E) = argmaxA (e / (l+1)mPj=1..m p(fj|eaj)) = argmaxA Pj=1..mp(fj|eaj) (IBM Model 1) • Again, use dynamic programming, Viterbi-like algorithm. • Recompute p(f|e) based on the best alignment • (only if you are inclined to do so; the “original” summed-over-all distribution might perform better). • Note: we have also got all Model 1 parameters. JHU CS 600.465/ Intro to NLP/Jan Hajic

More Related