1 / 21

Learning to recognize features of valid textual inferences

Learning to recognize features of valid textual inferences. Bill MacCartney Stanford University with Trond Grenager, Marie-Catherine de Marneffe, Daniel Cer, and Christopher D. Manning. The textual inference task. Does text T justify an inference to hypothesis H ?

dotty
Download Presentation

Learning to recognize features of valid textual inferences

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. Learning to recognize featuresof valid textual inferences Bill MacCartneyStanford University withTrond Grenager, Marie-Catherine de Marneffe,Daniel Cer, and Christopher D. Manning

  2. The textual inference task • Does text T justify an inference to hypothesis H? • An informal, intuitive notion of inference: not strict logic • Focus on local inference steps, not long chains of deduction • Emphasis onvariability of linguistic expression • Robust, accurate textual inference would enable: • Semantic search:H: lobbyists attempting to bribe U.S. legislators T: The A.P. named two more senators who received contributions engineeredby lobbyist Jack Abramoff in return for political favors. • Question answering: H: Who bought J.D. Edwards? T: Thanks to its recent acquisition of J.D. Edwards, Oracle will soon be able… • Customer email response • Relation extraction (database building) • Document summarization

  3. Textual inference as graph alignment • Many efforts have converged on this approach[Haghighi et al. 05, de Salvo Braz et al. 05] • Represent T & H as typed dependency graphs • Graph nodes = words of sentence • Graph edges = grammatical relations (subject, possessive, etc.) • Find least-cost alignment of H to (part of) T • Can H be (approximately) embedded within T? • Use locally-decomposable cost model • Lexical costs penalize aligning semantically unrelated words • Structural costs penalize aligning dissimilar subgraphs • Assume good alignment  valid inference

  4. reported nsubj ccomp killed CNN lost nsubjpass in nsubj in aux dobj troops were ambush soldiers lives ambush amod det nn dep poss several the thirteen their today’s Example: graph alignment ⊨ • Problems: non-monotonicity, non-locality, … T: CNN reported that thirteen soldiers lost their lives in today’s ambush. H: Several troops were killed in the ambush.

  5. ⊨ ? ? Weighted abduction models[Hobbs et al. 93, Moldovan et al. 03, Raina et al. 05] • Translate to FOL and try to prove H from T • Allow assumptions at various “costs” • Superficially, like using formal semantics & logic • Actually, analogous to graph-matching approach • FOL  dep’cy graphs; abduction costs  lexical match costs • Modulo use of additional axioms [Tatu et al. 06] T: Kidnappers released a Filipino hostage. H: A Filipino hostage was freed. e, a, b kidnappers(a)  release(e, a, b) Filipino(b)  hostage(b) f, x Filipino(x)  hostage(x) freed(f, x) released(p, q, r)  sfreed(s, r)  enables proof; costs $2.00

  6. / Problems with alignment models • Alignments are important, but… • Good alignment valid inference: • Assumption of upward monotonicity • Assumption of locality • Confounding of alignment and entailment

  7. Problem 1: non-monotonicity • In normal “upward monotone” contexts, generalizing aconcept preserves truth: T: SomeKorean historians believe the murals are of Korean origin. H: Somehistorians believe the murals are of Korean origin. • But not in “downward monotone” contexts: T: FewKorean historians doubt that Koguryo belonged to Korea. H: Fewhistorians doubt that Koguryo belonged to Korea. • Lots of constructs invert monotonicity! ⊨ ⊭ • explicit negation: not • restrictive quantifiers: no, few, at most n • negative or restrictive verbs: lack, fail, deny • preps & adverbs: without, except, only • comparatives and superlatives • antecedent of a conditional: if

  8. Problem 2: non-locality • To be tractable, alignment scoring must be local • But valid inference can hinge on non-local factors: ⊨ T1: The army confirmed that interrogators desecrated the Koran. H: Interrogators desecrated the Koran. T2: Newsweek retracted its report thatthe army had confirmed that interrogators desecrated the Koran. H: Interrogators desecrated the Koran. ⊭

  9. Problem 3:confounding alignment & inference • If alignment  entailment, lexical cost model must penalize e.g. antonyms, inverses: T: Stocks fell on worries that oil prices would rise this winter. H: Stock prices climbed. • But aligner will seek the best alignment: T: Stocks fell on worries that oil prices would rise this winter. H: Stock prices climbed. • Actually, we want the first alignment, and then a separate assessment of entailment! [cf. Marsi & Krahmer 05] must prevent this alignment maybe entailed?

  10. acquires acquires buys buys –1.28 –0.88 dobj dobj dobj dobj nsubj nsubj nsubj nsubj missiles missiles arms arms India India India India 0.00 –0.75 –0.53 Solution: three-stage architecture ⊨ T: India buys missiles. H: India acquires arms. 1. linguisticanalysis 3. features &classification 2. graphalignment yes tunedthreshold score = = no

  11. 1. Linguistic analysis • Typed dependencies from statistical parser [de Marneffeetal.06] • Collocations from WordNet (Bill hung_up the phone) • Statistical named entity recognizers[Finkeletal.05] • Canonicalization of quantity, date, and money expressions T: Kessler’s team conducted60,643[60,643]face-to-face interviews... H: Kessler’s team interviewedmore than 60,000[>60,000]adults... • Semantic role identification: PropBank roles[Toutanovaetal.05] • Coreference resolution: T: Sinceitsformation in 1948,Israel… H: Israel was established in 1948. • Hand-built: acronyms, country and nationality, factive verbs • TF-IDF scores ⊨ ⊨

  12. 2. Aligning dependency graphs • Beam search for least-cost alignment • Locally decomposable cost model • Can’t do Viterbi-style DP or heuristic search without this • Assessment of global features postponed to next stage • Lexical matching costs • Use lexical semantic relatedness scores derived from WordNet, Infomap, gazetteers, distributional similarity [Lin 98] • Do not penalize antonyms, inverses, alternatives… • Structural matching costs • Each edge in graph of H determines path in graph of T • Preserved edges get best score; longer paths score lower

  13. 3. Features of valid inferences • After alignment, extract features of inference • Look for global characteristics of valid and invalid inferences • Features embody crude semantic theories • Feature categories: adjuncts, modals, quantifiers, implicatives, antonymy, tenses, structure, explicit numbers & dates • Alignment score is also an important feature • Extracted features  statistical model  score • Can learn feature weights using logistic regression • Or, can use hand-tuned weights • (Score ≥ threshold) ?  prediction: yes/no • Threshold can be tuned

  14. Features: restrictive adjuncts • Does hypothesis add/drop a restrictive adjunct? • Adjunct is dropped: usually truth-preserving • Adjunct is added: suggests no entailment • But in a downward monotone context, this is reversed T: In all, Zerich bought $422 million worth of oil from Iraq, according to the Volcker committee. H: Zerich bought oil from Iraqduring the embargo. T: Zerich didn’t buy any oil from Iraq, according to the Volcker committee. H: Zerich didn’t buy oil from Iraqduring the embargo. • Generate features for add/drop, monotonicity ⊭ ⊨

  15. Features: modality ⊭ T: Sharon warns Arafatcould be targeted for assassination. H: Prime minister targeted for assassination.[RTE1-98] T: After the trial, Family Court found the defendant guilty of violating the order. H: Family Court cannot punish the guilty.[RTE1-515] ⊭ • Define 6 canonical modalities • Identify modalities of T & H: • Map T, H modality pairs to categorical features:

  16. Features: factives & implicatives T: Libya has tried, with limited success, to develop its own indigenous missile, and to extend the range of its aging SCUD force for many years under the Al Fatah and other missile programs. H: Libya hasdeveloped its own domestic missile program. • Evaluate governing verbs for implicativity class • Unknown: say, tell, suspect, try, … • Fact: know, acknowledge, ignore, … • True: manage to, … • False: fail to, forget to, … • Need to check for -monotone context here too • not try to winnot win, but not manage to win not win ⊭ ⊭ ⊨

  17. Evaluation: PASCAL RTE • Recognizing textual “entailment” [Dagan, Glickman, and Magnini 2005] • Does text T “entail” the truth of hypothesis H? T: Wal-Mart defended itself in court today against claims that its female employees were kept out of jobs in management because they are women. H: Wal-Mart was sued for sexual discrimination. • High inter-annotator agreement (~95%) • RTE1 (2005): 567 dev pairs, 800 test pairs ⊨

  18. Results & useful features Most useful features Positive • Added adjunct in  context • Pred-arg structure match • Modal: yes • Text is embedded in factive • Good alignment score Negative • Date inserted/mismatched • Pred-arg structure mismatch • Quantifier mismatch • Bad alignment score • Different polarity • Modal: no/don’t know *confidence-weighted score (standard RTE1 evaluation metric)

  19. What we have trouble with • Non-entailment is easier than entailment • Good at finding knock-out features • But, hard to be certain that we’ve considered everything • Lots of adjuncts, but which are restrictive? H: Maurice wassubsequently killed in Angola. • Multiword “lexical” semantics/world knowledge • We’re pretty good at synonyms, hyponyms, antonyms • But we aren’t good at recognizing multi-word equivalences T: David McCool took the money and decided to start Muzzy Lane in 2002. H: David McCoolis the founder ofMuzzy Lane. [RTE2-379] • Other teams (e.g. LCC) have done well with paraphrase models

  20. Conclusion • Alignment models promising, but flawed: • Assumption of monotonicity • Assumption of locality • Confounding of alignment and inference • Solution: align, then judge validity of inference • We extract global-level semantic features • Working from richly-annotated, aligned dependency graphs… not just word sequences • Features are designed to embody crude semantic theories • Still lots of room to improve…

  21. END

More Related