Elamur Research⌕
☰
home / papers / 2009 / An effective Discourse Parser that uses Rich Linguistic Information

An effective Discourse Parser that uses Rich Linguistic Information

Summary

The paper addresses discourse parsing: building a tree over a text's segments and labeling each internal node with a rhetorical relation. Most earlier parsers used only shallow lexical and syntactic signals. Subba and Di Eugenio test whether deeper linguistic knowledge helps. They work with a corpus of home-repair how-to instructions, manually split into EDUs and annotated with 26 informational relations. They also add VerbNet-based compositional semantics, derived largely automatically with the LCFLEX parser and a CoreLex noun lexicon.

Relations are classified with Inductive Logic Programming (the Aleph system). The background knowledge includes:

  • words and part-of-speech tags,
  • WordNet hypernymy and meronymy,
  • discourse cues,
  • segment similarity,
  • semantic predicates,
  • the internal structure of larger segments.

This relational learner significantly beats Decision Trees, RIPPER and Naive Bayes, and it produces readable first-order rules. The classifier drives a modified shift-reduce parser, which can attach new segments anywhere on the right frontier of the tree. The parser clearly outperforms a right-branching majority-class baseline. Adding semantics gives a statistically significant gain.

Key points

  • Corpus: 176 instructional documents with 5744 EDUs and 53,250 words. Full semantic representations were obtained for 3257 EDUs (56.7%).
  • Relation annotation: 5172 relations were labeled using 26 types grouped into 5 classes. The most frequent is preparation:act (24.46%). Inter-annotator agreement was κ = 0.66, rising to 0.78 when preparation:act and step1:step2 are merged. Only 43% of relations carry an explicit cue.
  • Sentence-level parsing (relation F-score): SR-ILP with semantics reached 63.06%, compared with 58.58% without semantics (p = 0.006) and 34.32% for the baseline.
  • Document-level parsing (relation F-score): performance falls to 35.44%, against 22.46% for the baseline. The authors attribute the drop to errors made at lower levels, which propagate upward, and they argue for a less greedy parsing strategy.
  • Evaluation setup: 151 documents were used for training and 25 for testing. Spans, nuclearity and relations were scored with Marcu's (2000) method.

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: Discourse analysis

Cited by

47 papers of our library cite it (of the 140673 whose bibliographies we read); 19 of them have a page here, 28 not yet.

Cite

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

@inproceedings{subba2009effective,
  title = {{An effective Discourse Parser that uses Rich Linguistic Information}},
  author = {Subba, Rajen and Eugenio, Barbara Di},
  booktitle = {Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2009},
  url = {https://aclanthology.org/N09-1064/}
}

References

26 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Asher, N., and Lascarides, A.: Logics of Conversation. Cambridge University Press, 2003. find
  2. Baldridge, J. and Lascarides, A.: Probabilistic Head-Driven Parsing for Discourse Structure In Proceedings of the Ninth Conference on Computational Natural Language Learning (CoNNL), Ann Arbor, 2005. find
  3. Brill, E.: Transformation-based error-driven learning and natural language processing: A case study in part-of-speech tagging. Computational Linguistics, 21(4):543565, 1995. find our card
  4. Buitelaar, P.: CoreLex: Systematic Polysemy and Underspecification. Ph.D. Thesis, Brandies University, 1998. find
  5. Carlson, L. D. M. and Okurowski., M. E.: Building a discourse-tagged corpus in the framework of rhetorical structure theory. In Current Directions in Discourse and Dialogue pages 85–112, 2003. find our card
  6. Cristea, D.: An Incremental Discourse Parser Architecture. In D. Christodoulakis (Ed.) Proceedings of the Second International Conference - Natural Language Processing - Patras, Greece, June 2000. find
  7. Forbes, K., Miltsakaki, E., R. P. A. S. A. J. and Webber., B.: D-ltag system - discourse parsing with a lexicalized tree adjoining grammar. Information Stucture, Discourse Structure and Discourse Semantics, ESS-LLI 2001. find
  8. Grosz, B. J. and Sidner, C. L.: Attention, intention and the structure of discourse. Computational Linguistics 12:175–204, 1988. find
  9. Hobbs, J. R.: On the coherence and structure of discourse. In Polyani, Livia editor, The Structure of Discourse, 1985. find
  10. Kipper, K., H. T. D. and Palmer., M.: Class-based construction of a verb lexicon. AAAI-2000, Proceedings of the Seventeenth National Conference on Artificial Intelligence, 2000. find
  11. Mann, W. and Thompson, S.: Rhetorical structure theory: Toward a functional theory of text organization. Text, 8(3):243–281, 1988. find our card
  12. Marcu, D.: Instructions for Manually Annotating the Discourse Structures of Texts. Technical Report, University of Southern California, 1999. find
  13. Marcu, D.: The theory and practice of discourse parsing and summarization. Cambridge, Massachusetts, London, England, MIT Press, 2000. find
  14. Moser, M. G., Moore, J. D., and Glendening, E.: Instructions for Coding Explanations: Identifying Segments, Relations and Minimal Units. University of Pittsburgh, Department of Computer Science, 1996. find
  15. Muggleton, S. H.: Inverse entailment and progol. In New Generation Computing Journal 13:245–286, 1995. find
  16. Polanyi, L., Culy, C., van den Berg, M. H. and Thione, G. L.: A Rule Based Approach to Discourse Parsing. Proceedings of the 5th SIGdial Workshop in Discourse And Dialogue. Cambridge, MA USA pp. 108-117., May 1, 2004. find our card
  17. Prasad, R., Dinesh, N., Lee, A., Miltsakaki, E., Robaldo, L., Joshi, A., and Webber, B.: The Penn Discourse Treebank 2.0. LREC, 2008. find our card
  18. Rosé, C. P.: A Syntactic Framework for Semantic Interpretation, Proceedings of the ESSLLI Workshop on Linguistic Theory and Grammar Implementation, 2000. find
  19. Sporleder, C. and Lascarides., A.: Exploiting linguistic cues to classify rhetorical relations. Recent Advances in Natural Language Processing, 2005. find
  20. Soricut, R. and Marcu., D.: Sentence level discourse parsing using syntactic and lexical information. Proceedings of the Human Language Technology and North American Assiciation for Computational Linguistics Conference, 2003. find our card
  21. Subba, R., Di Eugenio, B., E. T.: Building lexical resources for princpar, a large coverage parser that generates principled semantic representations. LREC, 2006. find
  22. Subba, R.: Discourse Parsing: A Relational Learning Approach Ph.D. Thesis, University of Illinois Chicago, December 2008. find
  23. Webber, B.: DLTAG: Extending Lexicalized TAG to Discourse. Cognitive Science 28:751-779, 2004. find
  24. Wellner, B., Pustejovsky, J., C. H. R. S. and Rumshisky., A.: Classification of discourse coherence relations: An exploratory study using multiple knowledge sources. In Proceedings of the 7th SIGDIAL Workshop on Discourse and Dialogue, 2006. find our card
  25. Williams, S. and Reiter, E.: A corpus analysis of discourse relations for natural language generation. Proceedings of Corpus Linguistics, pages 899–908, 2003. find
  26. Wolf, F. and Gibson, E.: Representing discourse coherence: A corpus-based analysis. Computational Linguistics 31(2):249–287, 2005. find our card

↑ back to top