Reza Bosaghzadeh & Nathan Schneider LS2 ~ 1 December 2008

1. Unsupervised Approaches to Sequence Tagging, Morphology Induction, and Lexical Resource Acquisition Reza Bosaghzadeh & Nathan Schneider LS2 ~ 1 December 2008

2. Unsupervised methods • Sequence labeling (Part-of-Speech tagging) • Morphology Induction • Lexical Resource acquisition . pronoun verb preposition det noun adverb She ran to the station quickly un-supervise-dlearn-ing

3. Part of Speech (POS) Tagging • Prototype-driven model • Cluster based on a few examples • (Haghighi and Klein 2006) • Contrastive estimation • Discount positive examples at cost of implicit negative example • (Smith and Eisner 2005)

4. Target Label Prototypes Prototype Driven taggingOverview Annotated Data Unlabeled Data Prototype List + (Haghighi and Klein 2006)

5. Size Restrict Terms Location Features Prototypes Information Extraction: Classified Ads Newly remodeled2 Bdrms/1 Bath, spacious upper unit, located in Hilltop Mall area. Walking distance to shopping, public transportation, schools and park.Paid water and garbage.No dogs allowed. Newly remodeled 2 Bdrms/1 Bath, spacious upper unit, located in Hilltop Mall area. Walking distance to shopping, public transportation, schools and park. Paid water and garbage. No dogs allowed. Prototype List (Haghighi and Klein 2006)

6. PUNC NN VBN CC JJ CD IN DET NNS IN NNP RB Prototypes English POS Newly remodeled 2Bdrms/1Bath,spacious upper unit,locatedin Hilltop Mallarea.Walkingdistance toshopping, public transportation,schoolsandpark.Paid water andgarbage. Nodogs allowed. Newly remodeled 2 Bdrms/1 Bath, spacious upper unit, located in Hilltop Mall area. Walking distance to shopping, public transportation, schools and park. Paid water and garbage. No dogs allowed. Prototype List (Haghighi and Klein 2006)

7. Where to get Prototypes • 3 prototypes per tag • Automatically extracted by frequency (Haghighi and Klein 2006)

8. Prototypes • Features are same as (Smith & Eisner 05) • Trigram tagger • Word type, suffixes up to length 3, contains hyphen, contains digit, initial capitalization Tie each word to its most similar prototype • Similarity (Schuetze 93) • SVD dimensionality reduction • cos() between context vectors (Haghighi and Klein 2006)

9. Prototypes – Language Variation Pros • Can get prototypes for popular languages • Doesn’t require tagging dictionary Cons • Needs tag set • Needs prototypes for each tag

10. Contrastive Estimation • Already discussed in class • Key Idea: • Mutating training examples often gives ungrammatical (negative) sentences • During training, shift probability mass from generated negative examples to given positive examples (Smith and Eisner 2005)

11. Unsupervised POS taggingCurrent results Best supervised results (CRFs): 99.5% !

12. Unsupervised methods • Sequence labeling (Part-of-Speech tagging) • Morphology Induction • Lexical Resource acquisition . pronoun verb preposition det noun adverb She ran to the station quickly un-supervise-dlearn-ing

13. Unsupervised Approaches to Morphology • Morphology refers to the internal structure of words • A morpheme is a minimal meaningful linguistic unit • Morpheme segmentation is the process of dividing words into their component morphemes unsupervised learning .

14. Unsupervised Approaches to Morphology • Morphology refers to the internal structure of words • A morpheme is a minimal meaningful linguistic unit • Morpheme segmentation is the process of dividing words into their component morphemes un-supervise-dlearn-ing • Word segmentation is the process of finding word boundaries in a stream of speech or text unsupervisedlearningofnaturallanguage

15. Unsupervised Approaches to Morphology • Morphology refers to the internal structure of words • A morpheme is a minimal meaningful linguistic unit • Morpheme segmentation is the process of dividing words into their component morphemes un-supervise-dlearn-ing • Word segmentation is the process of finding word boundaries in a stream of speech or text unsupervised learning of natural language

16. ParaMor: Monson et al. (2007, 2008) • Learns inflectional paradigms from raw text • Requires only the vocabulary of a corpus • Looks at word counts of substrings, and proposes (stem, suffix) pairings based on type frequency • 3-stage algorithm • Stage 1: Candidate paradigms based on frequencies • Stages 2-3: Refinement of paradigm set via merging and filtering • Paradigms can be used for morpheme segmentation or stemming

17. ParaMor: Monson et al. (2007, 2008) • A sampling of Spanish verb conjugations

18. ParaMor: Monson et al. (2007, 2008) • A proposed paradigm with stems {habl, bail} and suffixes {-ar, -o, -amos, -an}

19. ParaMor: Monson et al. (2007, 2008) • Same paradigm from previous slide, but with stems {habl, bail, compr}

20. ParaMor: Monson et al. (2007, 2008) • From just this list, other paradigm analyses (which happen to be incorrect) are possible

21. ParaMor: Monson et al. (2007, 2008) • Another possibility: stems {hab, bai}, suffixes {-lar, -lo, -lamos, -lan}

22. ParaMor: Monson et al. (2007, 2008) • Spurious segmentations—this paradigm doesn’t generalize to comprar(or most verbs)

23. ParaMor: Monson et al. (2007, 2008) • What if not all conjugations were in the corpus?

24. ParaMor: Monson et al. (2007, 2008) • We have two similar paradigms that we want to merge

25. ParaMor: Monson et al. (2007, 2008) • This amounts to smoothing, or “hallucinating” out-of-vocabulary items

26. ParaMor: Monson et al. (2007, 2008) • Heuristic-based, deterministic algorithm can learn inflectional paradigms from raw text • Paradigms can be used straightforwardly to predict segmentations • Combining the outputs of ParaMor and Morfessor (another system) won the segmentation task at MorphoChallenge 2008 for every language: English, Arabic, Turkish, German, and Finnish • Currently, ParaMor assumes suffix-based morphology

27. Goldwater et al. (2006; in submission) • Word segmentation results – comparison • See Narges & Andreas’s presentation for more on this model Goldwater Unigram DP Goldwater Bigram HDP

28. Multilingual morpheme segmentation Snyder & Barzilay (2008) • Considers parallel phrases and tries to find morpheme correspondences • Stray morphemes don’t correspond across languages • Abstract morphemes cross languages: (ar,er), (o, e), (amos,ons), (an,ent), (habl, parl)

29. Morphology papers: inputs & outputs

30. Unsupervised methods • Sequence labeling (Part-of-Speech tagging) • Morphology Induction • Lexical Resource acquisition . pronoun verb preposition det noun adverb She ran to the station quickly un-supervise-dlearn-ing

31. Lexical Resource Acquisition • Learning Bilingual Lexicons from Monolingual Corpora (Haghighi et al. 2008) • Narrative Event Chain Acquisition (Chambers and Jurafsky 2008) • Labeling semantic roles for verbs (Grenager and Manning 2006)

32. Learning Bilingual Lexicons from Monolingual Corpora Source Words s Target Words t Source Text Target Text Matching m nombre estado state world (Haghighi et al. 2008) política name mundo nation

33. Bilingual Lexicons: Data Representation estado state Orthographic Features Orthographic Features #st #es 1.0 1.0 tat sta 1.0 1.0 Source Text Target Text do# te# 1.0 1.0 Context Features Context Features world mundo 20.0 17.0 politics politica 10.0 5.0 Use CCA (Canonical Correlation Analysis) Trained with Hard EM to find best matching sociedad society 10.0 6.0 (Haghighi et al. 2008)

34. Feature Experiments • MCCA: Orthographic and context features Precision 4k EN-ES Wikipedia Articles (Haghighi et al. 2008)

35. Narrative events: Chambers & Jurafsky (2008) • Given a corpus, identifies related events that constitute a “narrative” and (when possible) predict their typical temporal ordering • E.g.: criminal prosecution narrative, with verbs: arrest, accuse, plead, testify, acquit/convict • Key insight: related events tend to share a participant in a document • The common participant may fill different syntactic/semantic roles with respect to verbs: arrest.object, accuse.object, plead.subject

36. Narrative events: Chambers & Jurafsky (2008) • A temporal classifier can reconstruct pairwise canonical event orderings, producing a directed graph for each narrative

37. Narrative events: Grenager & Manning (2006) • From dependency parses, a generative model predicts semantic roles corresponding to each verb’s arguments, as well as their syntactic realizations • PropBank-style: arg0, arg1, etc. per verb (do not necessarily correspond across verbs) • Learned syntactic patterns of the form: (subj=give.arg0, verb=give, np#1=give.arg2, np#2=give.arg1) or (subj=give.arg0, verb=give, np#2=give.arg1, pp_to=give.arg2) • Used for semantic role labeling

38. “Semanticity”: Our proposed scale of semantic richness • text < POS < syntax/morphology/alignments < coreference/semantic roles/temporal ordering < translations/narrative event sequences • We score each model’s inputs and outputs on this scale, and call the input-to-output increase “semantic gain” • Haghighi et al.’s bilingual lexicon induction wins in this respect, going from raw text to lexical translations

39. Robustness to language variation • About half of the papers we examined had English-only evaluations • We considered which techniques were most adaptable to other (esp. resource-poor) languages. Two main factors: • Reliance on existing tools/resourcesfor preprocessing (parsers, coreference resolvers, …) • Any linguistic specificityin the model (e.g. suffix-based morphology)

40. Summary We examined three areas of unsupervised NLP: • Sequence tagging: How can we predict POS (or topic) tags for words in sequence? • Morphology: How are words put together from morphemes (and how can we break them apart)? • Lexical resources: How can we identify lexical translations, semantic roles and argument frames, or narrative event sequences from text? In eight recent papers we found a variety of approaches, including heuristic algorithms, Bayesian methods, and EM-style techniques.

41. Questions? Thanks to Noah and Kevin for their feedback on the paper; Andreas and Narges for their collaboration on the presentations; and all of you for giving us your attention! subj=give.arg0verb=givenp#1=give.arg2np#2=give.arg1 • un-supervise-dlearn-ing