Elamur Research⌕
☰
home / papers / 2012 / Coreference Semantics from Web Features

Coreference Semantics from Web Features

Summary

Bansal and Klein tackle a weakness of coreference resolution: many hard decisions depend on world knowledge, such as knowing that "Obama" can be "the president", rather than on syntax. Structured knowledge bases cover only part of this knowledge. The authors instead draw shallow but broad semantic signals from the Google n-grams corpus. They start from the mention-pair model in Reconcile, replacing its averaged perceptron with a decision tree. For each pair of mention headwords, they then compute Web-count features in five groups:

  • plain co-occurrence
  • Hearst-pattern (hypernymy) co-occurrence
  • overlap of entity-based context seeds
  • matching of distributional cluster ids
  • substitution of candidate antecedents into the context around a pronoun

The features are tuned only on the ACE04 development set and reused without retuning on ACE05 and ACE05-ALL. They yield significant MUC and B³ gains over the decision-tree baseline and beat Haghighi and Klein (2010) on most settings. At publication, these were the best reported end-to-end results with system-detected mentions. The work shows that aggregated Web statistics can supply useful semantics cheaply.

Key points

  • Resources and implementation:
    • Web source: Google n-grams (Brants and Franz, 2006), covering n = 1–5 and built from roughly 1 trillion tokens and 95 billion sentences.
    • All queries are answered in a single pass using the trie-based hashing of Bansal and Klein (2011).
    • Cluster features rely on the clusters of Lin et al. (2010).
  • Baseline: Reconcile (over 80 core features) with Weka's J48 (C4.5) decision tree. This outperformed the default averaged perceptron, likely because trees capture non-linear feature interactions.
  • Feature construction: Co-occurrence counts allow 1–3 wildcards between the headwords. Counts are normalized by unigram frequencies (similar to scaled MI), then log10-binned.
    • Example: the pair (leader, president) has a count of 11383, versus 95 for (voter, president).
  • Significance: Gains are significant under the bootstrap test, except ACE05 with MUC (p < 0.12). ACE05-ALL shows larger gains at p < 0.001. Untuned B³ None on ACE05-ALL: P=69.9, R=65.9, F1=67.8.
  • Tree analysis: About 30% of the learned tree's nodes are Web features. Leaves on these features have only around 2.5% classification error. Corrected errors include pairs like "Iran ; the country" and "Barry Bonds ; the best baseball player".

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: Linguistics & languages

Topics: Coreference resolution

Cited by

21 papers of our library cite it (of the 140673 whose bibliographies we read); 15 of them have a page here, 6 not yet.

Cite

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

@inproceedings{bansal2012coreference,
  title = {{Coreference Semantics from Web Features}},
  author = {Bansal, Mohit and Klein, Dan},
  booktitle = {Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2012},
  url = {https://aclanthology.org/P12-1041/}
}

References

35 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Amit Bagga and Breck Baldwin. 1998. Algorithms for scoring coreference chains. In Proceedings of MUC-7 and LREC Workshop. find our card
  2. Mohit Bansal and Dan Klein. 2011. Web-scale features for full-scale parsing. In Proceedings of ACL. find
  3. Eric Bengtson and Dan Roth. 2008. Understanding the value of features for coreference resolution. In Proceedings of EMNLP. find our card
  4. Shane Bergsma and Dekang Lin. 2006. Bootstrapping path-based pronoun resolution. In Proceedings of COLING-ACL. find our card
  5. Shane Bergsma, Emily Pitler, and Dekang Lin. 2010. Creating robust supervised classifiers via web-scale ngram data. In Proceedings of ACL. find
  6. Thorsten Brants and Alex Franz. 2006. The Google Web 1T 5-gram corpus version 1.1. LDC2006T13. find
  7. Kenneth Ward Church and Patrick Hanks. 1989. Word association norms, mutual information, and lexicography. In Proceedings of ACL. find our card
  8. Hal Daumé III and Daniel Marcu. 2005. A large-scale exploration of effective global features for a joint entity detection and tracking model. In Proceedings of EMNLP. find
  9. B. Efron and R. Tibshirani. 1993. An introduction to the bootstrap. Chapman & Hall CRC. find
  10. Aria Haghighi and Dan Klein. 2009. Simple coreference resolution with rich syntactic and semantic features. In Proceedings of EMNLP. find our card
  11. Aria Haghighi and Dan Klein. 2010. Coreference resolution in a modular, entity-centered model. In Proceedings of NAACL-HLT. find our card
  12. Mark Hall, Eibe Frank, Geoffrey Holmes, Bernhard Pfahringer, Peter Reutemann, and Ian H. Witten. 2009. The WEKA data mining software: An update. SIGKDD Explorations, 11(1). find
  13. Zellig Harris. 1954. Distributional structure. Word, 10(23):146162. find
  14. Marti Hearst. 1992. Automatic acquisition of hyponyms from large text corpora. In Proceedings of COLING. find
  15. Hamidreza Kobdani, Hinrich Schutze, Michael Schiehlen, and Hans Kamp. 2011. Bootstrapping coreference resolution using word associations. In Proceedings of ACL. find our card
  16. Terry Koo, Xavier Carreras, and Michael Collins. 2008. Simple semi-supervised dependency parsing. In Proceedings of ACL. find our card
  17. Dekang Lin and Xiaoyun Wu. 2009. Phrase clustering for discriminative learning. In Proceedings of ACL. find
  18. Dekang Lin, Kenneth Church, Heng Ji, Satoshi Sekine, David Yarowsky, Shane Bergsma, Kailash Patil, Emily Pitler, Rachel Lathbury, Vikram Rao, Kapil Dalwani, and Sushant Narsale. 2010. New tools for web-scale n-grams. In Proceedings of LREC. find
  19. Katja Markert and Malvina Nissim. 2005. Comparing knowledge sources for nominal anaphora resolution. Computational Linguistics, 31(3):367–402. find our card
  20. Vincent Ng and Claire Cardie. 2002. Improving machine learning approaches to coreference resolution. In Proceedings of ACL. find our card
  21. Vincent Ng. 2010. Supervised noun phrase coreference research: The first fifteen years. In Proceedings of ACL. find our card
  22. NIST. 2004. The ACE evaluation plan. In NIST. find
  23. E.W. Noreen. 1989. Computer intensive methods for hypothesis testing: An introduction. Wiley, New York. find
  24. Massimo Poesio, Rahul Mehta, Axel Maroudas, and Janet Hitzeman. 2004. Learning to resolve bridging references. In Proceedings of ACL. find our card
  25. J. R. Quinlan. 1993. C4.5: Programs for machine learning. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA. find our card
  26. Altaf Rahman and Vincent Ng. 2009. Supervised models for coreference resolution. In Proceedings of EMNLP. find our card
  27. Altaf Rahman and Vincent Ng. 2011. Coreference resolution with world knowledge. In Proceedings of ACL. find our card
  28. Deepak Ravichandran, Patrick Pantel, and Eduard Hovy. 2005. Randomized algorithms and NLP: Using locality sensitive hash functions for high speed noun clustering. In Proceedings of ACL. find
  29. Marta Recasens and Eduard Hovy. 2010. Coreference resolution across corpora: Languages, coding schemes, and preprocessing information. In Proceedings of ACL. find our card
  30. Wee Meng Soon, Hwee Tou Ng, and Daniel Chung Yong Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational Linguistics, 27(4):521–544. find our card
  31. Veselin Stoyanov, Nathan Gilbert, Claire Cardie, and Ellen Riloff. 2009. Conundrums in noun phrase coreference resolution: Making sense of the state-of-the-art. In Proceedings of ACL/IJCNLP. find our card
  32. Veselin Stoyanov, Claire Cardie, Nathan Gilbert, Ellen Riloff, David Buttler, and David Hysom. 2010. Reconcile: A coreference resolution research platform. In Technical report, Cornell University. find
  33. Marc Vilain, John Burger, John Aberdeen, Dennis Connolly, and Lynette Hirschman. 1995. A model-theoretic coreference scoring scheme. In Proceedings of MUC-6. find our card
  34. Xiaofeng Yang and Jian Su. 2007. Coreference resolution using semantic relatedness information from automatically discovered patterns. In Proceedings of ACL. find our card
  35. Xiaofeng Yang, Jian Su, and Chew Lim Tan. 2005. Improving pronoun resolution using statistics-based semantic compatibility information. In Proceedings of ACL. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/bansal-2012-coreference-semantics-web-features/