Elamur Research⌕
☰
home / papers / 2007 / ISP: Learning Inferential Selectional Preferences

ISP: Learning Inferential Selectional Preferences

Paper · 2007 · Patrick Pantel, Rahul Bhagat, Bonaventura Coppola et al.

Summary

Automatically harvested inference rules (e.g., from DIRT) are often applied in contexts where they do not hold. A rule linking "charged by" to "announced the arrest of" works for criminal suspects but fails when the argument is a bank account. The authors present ISP, a set of methods that learn which argument types make a rule applicable, called inferential selectional preferences. ISP uses these preferences to reject inferences that do not fit.

ISP gathers corpus statistics on how binary relations are instantiated and maps each argument to semantic classes from WordNet or CBC clusters. It scores these preferences in two ways:

  • Joint models (JRM/JIM) treat the two arguments together and score class pairs with PMI.
  • Independent models (IRM/IIM) treat each argument separately, which helps with sparse data.

A rule's preferences are formed by intersecting the preferences of its two sides. The authors evaluated ISP on DIRT rules. The most permissive filter clearly beat the baselines, and the strict variants gave high specificity. This makes noisy rule collections more usable for question answering and textual entailment.

Key points

  • Variants and tuning: The variants are ISP.JIM, ISP.IIM.∧ (both arguments must match) and ISP.IIM.∨ (one argument suffices). The authors tuned them across 180 parameter configurations on a development set, then tested on a held-out set.
  • Best accuracy: ISP.IIM.∨ achieved the highest accuracy, about 0.62 at threshold τ=30%. It was significantly better than all three baselines, and its accuracy stayed within 0.05 of that peak across all τ values.
  • High-specificity option: Selected for at least 90% specificity, ISP.JIM recovers roughly 17% of correct inferences with a false-positive rate of about 12%.
  • WordNet vs. CBC: CBC classes outperformed the authors' simple WordNet-based classes, largely because CBC covers more words, especially proper names. Arguments that belong to no class cause the inference to be rejected.
  • Main remaining error: Antonym-like DIRT rules (e.g., "rejected in" ⇒ "accepted in") slip through. Such opposite relations share the same argument types, so ISP learns preferences that accept them.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

Cite

Cite as Patrick Pantel, Rahul Bhagat, Bonaventura Coppola, Timothy Chklovski and Eduard Hovy. 2007. ISP: Learning Inferential Selectional Preferences. Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceedings of the Main Conference.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{pantel2007isp,
  title = {{ISP: Learning Inferential Selectional Preferences}},
  author = {Patrick Pantel, Rahul Bhagat, Bonaventura Coppola, Timothy Chklovski and Eduard Hovy},
  booktitle = {Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceedings of the Main Conference},
  year = {2007},
  url = {https://aclanthology.org/N07-1071/}
}

References

20 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Barzilay, R.; and McKeown, K.R. 2001.Extracting Paraphrases from a Parallel Corpus. In Proceedings of ACL 2001. pp. 50–57. Toulose, France.
  2. Baker, C.F.; Fillmore, C.J.; and Lowe, J.B. 1998. The Berkeley FrameNet Project. In Proceedings of COLING/ACL 1998. pp. 86-90. Montreal, Canada.
  3. Cover, T.M. and Thomas, J.A. 1991. Elements of Information Theory. John Wiley & Sons. find
  4. Fellbaum, C. 1998. WordNet: An Electronic Lexical Database. MIT Press. find
  5. Harabagiu, S.; and Hickl, A. 2006. Methods for Using Textual Entailment in Open-Domain Question Answering. In Proceedings of ACL 2006. pp. 905-912. Sydney, Australia.
  6. Katz, J.; and Fodor, J.A. 1963. The Structure of a Semantic Theory. Language, vol 39. pp.170–210. find
  7. Lenat, D. 1995. CYC: A large-scale investment in knowledge infrastructure. Communications of the ACM, 38(11):33–38. find
  8. Levin, B. 1993. English Verb Classes and Alternations: A Preliminary Investigation. University of Chicago Press, Chicago, IL. find
  9. Light, M. and Greiff, W.R. 2002. Statistical Models for the Induction and Use of Selectional Preferences. Cognitive Science,26:269–281. find
  10. Lin, D. 1993. Parsing Without OverGeneration. In Proceedings of ACL-93. pp. 112-120. Columbus, OH. find
  11. Lin, D. and Pantel, P. 2001. Discovery of Inference Rules for Question Answering. Natural Language Engineering 7(4):343-360. find
  12. Moldovan, D.I.; Clark, C.; Harabagiu, S.M.; Maiorano, S.J. 2003. COGEX: A Logic Prover for Question Answering. In Proceedings of HLT-NAACL-03. pp. 87-93. Edmonton, Canada. find
  13. Pantel, P. and Lin, D. 2002. Discovering Word Senses from Text. In Proceedings of KDD-02. pp. 613-619. Edmonton, Canada. find
  14. Resnik, P. 1996. Selectional Constraints: An Information-Theoretic Model and its Computational Realization. Cognition, 61:127–159. find
  15. Romano, L.; Kouylekov, M.; Szpektor, I.; Dagan, I.; Lavelli, A. 2006. Investigating a Generic Paraphrase-Based Approach for Relation Extraction. In EACL-2006. pp. 409-416. Trento, Italy. find
  16. Rosch, E. 1978. Human Categorization. In E. Rosch and B.B. Lloyd (eds.) Cognition and Categorization. Hillsdale, NJ: Erlbaum. find
  17. Siegel, S. and Castellan Jr., N. J. 1988. Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill. find
  18. Szpektor, I.; Tanev, H.; Dagan, I.; and Coppola, B. 2004. Scaling web-based acquisition of entailment relations. In Proceedings of EMNLP 2004. pp. 41-48. Barcelona,Spain.
  19. Wilks, Y.; and Fass, D. 1992. Preference Semantics: a family history. Computing and Mathematics with Applications, 23(2). A shorter version in the second edition of the Encyclopedia of Artificial Intelligence, (ed.) S. Shapiro. find
  20. Zanzotto, F.M.; Pennacchiotti, M.; Pazienza, M.T. 2006. Discovering Asymmetric Entailment Relations between Verbs using Selectional Preferences. In COLING/ACL-06. pp. 849-856. Sydney, Australia. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pantel-2007-isp-learning-inferential-selectional-preferences/