Elamur Research⌕
☰
home / papers / 2008 / A Global Joint Model for Semantic Role Labeling

A Global Joint Model for Semantic Role Labeling

Summary

Most semantic role labeling (SRL) systems label each parse-tree constituent separately. This ignores a basic linguistic fact: a verb's arguments form a frame whose members constrain one another. For example, an agent rarely appears twice, and one argument's role can depend on how the others are realized. Toutanova, Haghighi and Manning build a discriminative log-linear joint model over whole argument frames. Formally it is a densely connected CRF with no Markov restrictions. Exact search over all labelings is infeasible, so a local SRL classifier first proposes n-best complete labelings, which the joint model then re-ranks. To limit overfitting, the authors use only a small set of linguistically motivated joint features. These include features of the full label sequence and features that link one argument's label to the syntactic properties of other arguments. On Propbank, the joint model clearly beats a strong local baseline, with gold and automatic parses alike. On CoNLL 2005 data, it surpasses the shared task's winning system. The work shows that statistical modeling of argument interdependencies pays off. It also shows that considering several parser outputs helps counter parsing noise.

Key points

  • On Propbank with gold-standard parse trees, the joint model lowers the local model's error rate by 24.1% when every argument is counted, and by 36.8% when only core arguments (ARG0–ARG5) are counted.
  • With automatic parses, the error reductions are smaller: 8.3% on all arguments and 10.3% on core arguments.
  • Inference uses Collins-style re-ranking of the local model's n-best assignments. The joint constraints are learned as soft statistical preferences, unlike the hard, hand-written ILP constraints of Punyakanok et al.
  • On the CoNLL 2005 data, choosing among the top five Charniak parses yields an F-Measure of 80.32 on WSJ Test (winning system: 79.44) and 68.81 on Brown Test (winning system: 67.75).
  • These scores exceed the authors' original shared-task submission (78.45 on WSJ Test) because of two fixes: continuation labels are now added only to core arguments, and a bug in how forward quotes reached the Charniak parser was corrected.

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: Natural language processing

Topics Semantic role labeling

Institutions

🎓Stanford UniversityStanford · United States248 works · 214,804 citations🎓University of California, BerkeleyBerkeley · United States107 works · 205,102 citations🏢MicrosoftRedmond · United States193 works · 110,465 citations

Cite

Cite as Kristina Toutanova, Aria Haghighi and Christopher D. Manning. 2008. A Global Joint Model for Semantic Role Labeling. Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1162/coli.2008.34.2.161

@article{toutanova2008global,
  title = {{A Global Joint Model for Semantic Role Labeling}},
  author = {Kristina Toutanova, Aria Haghighi and Christopher D. Manning},
  journal = {Computational Linguistics},
  year = {2008},
  doi = {10.1162/coli.2008.34.2.161},
  url = {https://doi.org/10.1162/coli.2008.34.2.161}
}

References

27 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Baker, Collin, Charles Fillmore, and John Lowe. 1998. The Berkeley Framenet project. In Proceedings of COLING-ACL, pages 86–90, San Francisco, CA. our card
  2. Carreras, Xavier and Luı́s Màrquez. 2004. Introduction to the CoNLL-2004 shared task: Semantic role labeling. In Proceedings of CoNLL, pages 89–97, Boston, MA. find
  3. Carreras, Xavier and Luı́s Màrquez. 2005. Introduction to the CoNLL-2005 shared task: Semantic role labeling. In Proceedings of CoNLL, pages 152–164, Ann Arbor, MI. our card
  4. Charniak, Eugene. 2000. A maximum-entropy-inspired parser. In Proceedings of NAACL, pages 132–139, Seattle, WA. our card
  5. Charniak, Eugene and Mark Johnson. 2005. Coarse-to-fine n-best parsing and MaxEnt discriminative reranking. In Proceedings of ACL, pages 173–180, Ann Arbor, MI. our card
  6. Cohn, Trevor and Philip Blunsom. 2005. Semantic role labelling with tree conditional random fields. In Proceedings of CoNLL, pages 169–172, Ann Arbor, MI. find
  7. Collins, Michael. 2000. Discriminative reranking for natural language parsing. In Proceedings of ICML, pages 175–182, Stanford, CA. our card
  8. Finkel, Jenny, Christopher Manning, and Andrew Ng. 2006. Solving the problem of cascading errors: Approximate bayesian inference for linguistic annotation pipelines. In Proceedings of EMNLP, pages 618–626, Sydney, Australia. find
  9. Gildea, Daniel and Daniel Jurafsky. 2002. Automatic labeling of semantic roles. Computational Linguistics, 28(3):245–288. our card
  10. Hacioglu, Kadri. 2004. A lightweight semantic chunking model based on tagging. In Proceedings of HLT-NAACL: Short Papers, pages 145–148, Boston, MA. find
  11. Haghighi, Aria, Kristina Toutanova, and Christopher D. Manning. 2005. A joint model for semantic role labeling. In Proceedings of CoNLL, pages 173–176, Ann Arbor, MI. find
  12. Lafferty, John, Andrew McCallum, and Fernando Pereira. 2001. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings of ICML, pages 282–289, Williamstown, MA. find
  13. Levy, Roger and Chris Manning. 2004. Deep dependencies from context-free statistical parsers: correcting the surface dependency approximation. In Proceedings of ACL, pages 327–334, Barcelona, Spain. find
  14. Màrquez, Luı́s, Mihai Surdeanu, Pere Comas, and Jordi Turmo. 2005. A robust combination strategy for semantic role labeling. In Proceedings of EMNLP, pages 644–651, Vancouver, Canada. find
  15. McCallum, Andrew, Dayne Freitag, and Fernando Pereira. 2000. Maximum entropy Markov models for information extraction and segmentation. In Proceedings of ICML, pages 591–598, Stanford, CA. find
  16. Palmer, Martha, Dan Gildea, and Paul Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. Computational Linguistics, 31(1):71–105. our card
  17. Pradhan, Sameer, Kadri Hacioglu, Valerie Krugler, Wayne Ward, James Martin, and Dan Jurafsky. 2005. Support vector learning for semantic argument classification. Machine Learning Journal, 60(1):11–39. find
  18. Pradhan, Sameer, Wayne Ward, Kadri Hacioglu, James Martin, and Dan Jurafsky. 2004. Shallow semantic parsing using support vector machines. In Proceedings of HLT-NAACL, pages 233–240, Boston, MA. our card
  19. Pradhan, Sameer, Wayne Ward, Kadri Hacioglu, James Martin, and Daniel Jurafsky. 2005. Semantic role labeling using different syntactic views. In Proceedings of ACL, pages 581–588, Ann Arbor, MI. find
  20. Punyakanok, Vasin and Dan Roth. 2001. The use of classifiers in sequential inference. In Proceedings of NIPS, pages 995–1001, Vancouver, Canada. find
  21. Punyakanok, Vasin, Dan Roth, and Wen-tau Yih. 2005. The necessity of syntactic parsing for semantic role labeling. In Proceedings of IJCAI, pages 1117–1123, Acapulco, Mexico. find
  22. Punyakanok, Vasin, Dan Roth, Wen-tau Yih, Dav Zimak, and Yuancheng Tu. 2004. Semantic role labeling via generalized inference over classifiers. In Proceedings of CoNLL, pages 130–133, Boston, MA. find
  23. Roland, Douglas and Daniel Jurafsky. 2002. Verb sense and verb subcategorization probabilities. In Paola Merlo and Suzanne Stevenson, editors, The Lexical Basis of Sentence Processing: Formal, Computational, and Experimental Issues. John Benjamins, Amsterdam, pages 325–345. find
  24. Sha, Fei and Fernando Pereira. 2003. Shallow parsing with conditional random fields. In Proceedings of HLT-NAACL, pages 134–141, Edmonton, Canada. find
  25. Surdeanu, Mihai, Sanda Harabagiu, John Williams, and Paul Aarseth. 2003. Using predicate-argument structures for information extraction. In Proceedings of ACL, pages 8–15, Sapporo, Japan. our card
  26. Xue, Nianwen and Martha Palmer. 2004. Calibrating features for semantic role labeling. In Proceedings of EMNLP, pages 88–94, Barcelona, Spain. find
  27. Yi, Szu-ting and Martha Palmer. 2005. The integration of syntactic parsing and semantic role labeling. In Proceedings of CoNLL, pages 237–240, Ann Arbor, MI. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/toutanova-2008-global-joint-model-semantic-role/