Elamur Research⌕
☰
home / papers / 2013 / The Life and Death of Discourse Entities: Identifying Singleton Mentions

The Life and Death of Discourse Entities: Identifying Singleton Mentions

Paper · 2013 · Marta Recasens, Marie-Catherine de Marneffe and Christopher Potts

Summary

Most entities introduced in a text are mentioned only once. The authors ask whether such one-off mentions (singletons) can be told apart from mentions that join a coreference chain, using only properties of the noun phrase and its context. They use the CoNLL-2012 Shared Task portion of OntoNotes v5.0. Because OntoNotes leaves singletons unannotated, every NP outside a chain is labelled singleton. On this data they fit a binary logistic regression, the lifespan model, with three feature groups grounded in linguistic theory:

  • the mention's internal morphosyntax;
  • its grammatical role;
  • its semantic environment (negation, modals, attitude predicates) and that environment's interactions with mention type.

The model reaches 78% accuracy, and its coefficients largely agree with predictions from dynamic semantics and Centering Theory. The authors then let the Stanford rule-based coreference system skip mention pairs that the model confidently marks as singleton. This yields small but statistically significant F1 gains on CoNLL-2012 test data. The work shows that theory-driven, interpretable features can reliably flag entities that will never be referred to again.

Key points

  • Data: CoNLL-2012 English data drawn from OntoNotes v5.0 (about 1.6M words). Singletons make up 56% of mentions in the dev set and include non-referential NPs.
  • The STANDARD model (threshold 0.5) reaches 78% accuracy. The CONFIDENT variant only predicts singleton below Pr 0.2 or coreferent above 0.8, which raises singleton precision to 89.6 while recall falls to 50.5.
  • Pronouns, proper nouns and singular animates favour coreference. Quantified NPs, indefinites, ORDINAL/PERCENT/QUANTITY entities, and mentions under negation or modals favour singletons. Indefinites inside negation are pushed even further toward singleton status.
  • The model was integrated into the Stanford sieve system (Lee et al., 2011), with predictions ignored for named entities. Gains of 0.5–1 F1 appear on MUC, CEAF-φ3, CEAF-φ4 and the CoNLL score (60.65 → 61.13).
  • An apparent drop in B3 was traced to a bug in the CoNLL-2012 scorer's mention alignment. A scorer corrected to follow Cai and Strube (2010) shows B3 improving.

Interesting ideas

  • I1. The model only acts on predictions inside a high-confidence band and abstains otherwise, so it can veto candidates in a precision-ordered pipeline without causing much damage.
  • I2. Semantic scope is not available directly, so the paper approximates it with syntactic scope in dependency trees to decide whether a mention sits under negation, a modal or an attitude verb. This is a cheap and explainable stand-in for logical scope.
  • I3. Reported content under attitude verbs such as "said" behaves like main-clause material in news-style text: it supports later reference rather than blocking it. This suggests treating reporting verbs as source markers instead of as scope islands.
  • I4. Each feature is an interpretable regression coefficient, and the signs are checked against what linguistic theory predicts, so the classifier doubles as a test of the theory.
  • I5. When a metric moved the wrong way, the authors traced the cause to the scorer itself and re-scored with a corrected implementation. They did not simply accept the number.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): A singleton detector in prag's coreference step could remove confidently "dead" mentions from the antecedent candidate lists, while uncertain mentions pass through untouched.
  • I2 (world): Entities born under negation or a modal in the UD tree could be stored as hypothetical, so a later "it" is not resolved to an object that was never actually asserted to exist.
  • I3 (world): Content embedded under reporting verbs could enter the text's world as asserted facts tagged with their source, keeping their entities available for coreference and QA.
  • I4 (learn): Mova's learned components could be audited the same way, flagging any feature whose learned sign contradicts a stated linguistic expectation as a candidate bug or data problem.
  • I5 (learn): Evaluation could include a check that the scorer behaves sensibly on small hand-built cases, such as spurious links to mentions absent from gold, before any score change is trusted.

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

 Institutions

🏢GoogleMountain View · United States353 works · 364,806 citations🎓Stanford UniversityStanford · United States310 works · 284,480 citations

📚 Shelves 4 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works

🧭 Topics 2 Coreference resolution · 550 works Discourse analysis · 394 works

Cite

Cite as Marta Recasens, Marie-Catherine de Marneffe and Christopher Potts. 2013. The Life and Death of Discourse Entities: Identifying Singleton Mentions. Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies.

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

@inproceedings{recasens2013life,
  title = {{The Life and Death of Discourse Entities: Identifying Singleton Mentions}},
  author = {Marta Recasens, Marie-Catherine {de Marneffe} and Christopher Potts},
  booktitle = {Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
  year = {2013},
  url = {https://aclanthology.org/N13-1071/}
}

References

44 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Judith Aissen. 1997. On the syntax of obviation. Language, 73(4):705–750. ↗
  2. Amit Bagga and Breck Baldwin. 1998. Algorithms for scoring coreference chains. In Proceedings of the LREC 1998 Workshop on Linguistic Coreference, pages 563–566. our card
  3. C. L. Baker. 1970. Double negatives. Linguistic Inquiry, 1(2):169–186. find
  4. David Beaver. 2004. The optimization of discourse anaphora. Linguistics and Philosophy, 27(1):3–56. ↗
  5. Maria Bittner. 2001. Surface composition as bridging. Journal of Semantics, 18(2):127–177. ↗
  6. Jie Cai and Michael Strube. 2010. Evaluation metrics for end-to-end coreference resolution systems. In Proceedings of SIGDIAL 2010, pages 28–36. our card
  7. Wallace L. Chafe. 1976. Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View. In Charles N. Li, editor, Subject and Topic, pages 25–55. Academic Press, New York. find
  8. Marie-Catherine de Marneffe, Bill MacCartney, and Christopher D. Manning. 2006. Generating typed dependency parses from phrase structure parses. In Proceedings of LREC 2006. our card
  9. Marie-Catherine de Marneffe, Christopher D. Manning, and Christopher Potts. 2012. Did it happen? The pragmatic complexity of veridicality assessment. Computational Linguistics, 38(2):301–333. our card
  10. Pascal Denis and Jason Baldridge. 2009. Global joint models for coreference resolution and named entity classification. Procesamiento del Lenguaje Natural, 42:87–96. find
  11. Eraldo Fernandes, Cı́cero dos Santos, and Ruy Milidiú. 2012. Latent structure perceptron with feature induction for unrestricted coreference resolution. In Proceedings of CoNLL-2012: Shared Task, pages 41–48. our card
  12. Anastasia Giannakidou. 1999. Affective dependencies. Linguistics and Philosophy, 22(4):367–421. find
  13. Jeroen Groenendijk and Martin Stokhof. 1991. Dynamic predicate logic. Linguistics and Philosophy, 14(1):39– 100. find
  14. Barbara J. Grosz, Aravind K. Joshi, and Scott Weinstein. 1995. Centering: A framework for modeling the local coherence of discourse. Computational Linguistics, 21(2):203–225. Jesse A. Harris and Christopher Potts. Perspective-shifting with appositives and expressives. Linguistics and Philosophy, 32(6):523–552. our card
  15. Irene Heim. 1982. The Semantics of Definite and Indefinite Noun Phrases. Ph.D. thesis, UMass Amherst. ↗
  16. Irene Heim. 1992. Presupposition projection and the semantics of attitude verbs. Journal of Semantics, 9(2):183–221. ↗
  17. Jerry R. Hobbs. 1979. Coherence and coreference. Cognitive Science, 3(1):67–90. ↗
  18. Eduard Hovy, Mitchell Marcus, Martha Palmer, Lance Ramshaw, and Ralph Weischedel. 2006. OntoNotes: The 90% solution. In Proceedings of HLT-NAACL 2006, pages 57–60. find
  19. Michael Israel. 1996. Polarity sensitivity as lexical semantics. Linguistics and Philosophy, 19(6):619–666. ↗
  20. Michael Israel. 2001. Minimizers, maximizers, and the rhetoric of scalar reasoning. Journal of Semantics, 18(4):297–331. ↗
  21. Hans Kamp. 1981. A theory of truth and discourse representation. In Jeroen Groenendijk, Theo M. V. Janssen, and Martin Stockhof, editors, Formal Methods in the Study of Language, pages 277–322. Mathematical Centre, Amsterdam. find
  22. Lauri Karttunen. 1973. Presuppositions and compound sentences. Linguistic Inquiry, 4(2):169–193. find
  23. Lauri Karttunen. 1976. Discourse referents. In James D. McCawley, editor, Syntax and Semantics, volume 7: Notes from the Linguistic Underground, pages 363– 385. Academic Press, New York. find
  24. William A. Ladusaw. 1980. On the notion ‘affective’ in the analysis of negative polarity items. Journal of Linguistic Research, 1(1):1–16. find
  25. Heeyoung Lee, Yves Peirsman, Angel Chang, Nathanael Chambers, Mihai Surdeanu, and Dan Jurafsky. 2011. Stanford’s multi-pass sieve coreference resolution system at the CoNLL-2011 Shared Task. In Proceedings of CoNLL-2011: Shared Task, pages 28–34. our card
  26. Xiaoqiang Luo. 2005. On coreference resolution performance metrics. In Proceedings of HLT-EMNLP 2005, pages 25–32. our card
  27. Barbara H. Partee. 1987. Noun phrase interpretation and type-shifting principles. In Jeroen Groenendijk, Dick de Jong, and Martin Stokhof, editors, Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers, pages 115–143. Foris Publications, Dordrecht. ↗
  28. Sameer Pradhan, Alessandro Moschitti, Nianwen Xue, Olga Uryupina, and Yuchen Zhang. 2012. CoNLL-2012 Shared Task: Modeling multilingual unrestricted coreference in OntoNotes. In Proceedings of EMNLP and CoNLL-2012: Shared Task, pages 1–40. our card
  29. Ellen Prince. 1981a. On the inferencing of indefinite ‘this’ NPs. In Bonnie Lynn Webber, Ivan Sag, 2009. and Aravind Joshi, editors, Elements of Discourse Understanding, pages 231–250. Cambridge University Press, Cambridge. find
  30. Ellen F. Prince. 1981b. Toward a taxonomy of given– new information. In Peter Cole, editor, Radical Pragmatics, pages 223–255. Academic Press, New York. find
  31. R Core Team, 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. find
  32. Marta Recasens and Eduard Hovy. 2009. A deeper look into features for coreference resolution. In Sobha Lalitha Devi, António Branco, and Ruslan Mitkov, editors, Anaphora Processing and Applications, volume 5847 of Lecture Notes in Computer Science, pages 29– 42. Springer. find
  33. Marta Recasens and Eduard Hovy. 2011. BLANC: Implementing the Rand index for coreference evaluation. Natural Language Engineering, 17(4):485–510. ↗
  34. Craige Roberts. 1990. Modal Subordination, Anaphora, and Distributivity. Garland, New York. find
  35. Craige Roberts. 1996. Anaphora in intensional contexts. In Shalom Lappin, editor, The Handbook of Contemporary Semantic Theory, pages 215–246. Blackwell Publishers, Oxford. find
  36. Johan Rooryck. 2001. Evidentiality, Part II. Glot International, 5(5):161–168. find
  37. Roger Schwarzschild. 2002. Singleton indefinites. Journal of Semantics, 19(3):289–314. find
  38. Mandy Simons. 2007. Observations on embedding verbs, evidentiality, and presupposition. Lingua, 117(6):1034–1056. ↗
  39. Ton van der Wouden. 1997. Negative Contexts: Collocation, Polarity and Multiple Negation. Routledge, London and New York. find
  40. Marc Vilain, John Burger, John Aberdeen, Dennis Connolly, and Lynette Hirschman. 1995. A model-theoretic coreference scoring scheme. In Proceedings of MUC-6, pages 45–52. our card
  41. Marilyn A. Walker, Aravind K. Joshi, and Ellen F. Prince, editors. 1997. Centering in Discourse. Oxford University Press. find
  42. Marilyn A. Walker, Aravind K. Joshi, and Ellen F. Prince. 1998. Centering in naturally-occurring discourse: An overview. In Marilyn A. Walker, Aravind K. Joshi, and Ellen F. Prince, editors, Centering Theory in Discourse, pages 1–28, Oxford. Clarendon Press. ↗
  43. Linton Wang, Eric McCready, and Nicholas Asher. 2006. Information dependency in quantificational subordination. In Klaus von Heusinger and Ken Turner, editors, Where Semantics Meets Pragmatics, pages 267–304. Elsevier Science, Amsterdam. ↗
  44. Gregory Ward and Betty Birner. 2004. Information structure and non-canonical syntax. In Laurence R. Horn and Gregory Ward, editors, The Handbook of Pragmatics, pages 153–174. Blackwell, Oxford. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/recasens-2013-life-death-discourse-entities-identifying/