Elamur ResearchβŒ•
☰
home / papers / 2008 / Easily Identifiable Discourse Relations

Easily Identifiable Discourse Relations

Summary

Automatically recognizing discourse relations (causal, contrastive, temporal and so on) between text units is hard. The difficulty depends on how often a relation has no overt marker, and on how ambiguous the markers are when they do appear. The authors measured both on the Penn Discourse Tree Bank (PDTB), a manual annotation layer over the 1-million-word WSJ corpus of the Penn Treebank. It labels explicit relations, which have a connective, and implicit relations, which have none, using four top-level sense classes: Comparison, Contingency, Temporal and Expansion. They found that most connectives strongly favor a single sense. A decision tree that uses only connective-presence features reaches 93.09% accuracy on four-way classification of explicit relations and 74.74% on all relations. That makes it a demanding baseline for more sophisticated systems. A χ² analysis of neighboring relations shows that some relation pairs appear next to each other more often than chance would predict. So easy explicit relations could help label nearby implicit ones through sequence-level modeling.

Key points

  • PDTB size: 18,459 explicit relations (19,458 sense labels, from 100 connective types) and 16,224 implicit relations (16,584 senses).
  • How often relations are explicit: Temporal relations have a connective about 80% of the time and Comparison about 69%. Contingency is roughly half explicit (about 47%), and Expansion, the largest class, is mostly implicit (about 42% explicit).
  • Connective ambiguity: Connectives appear in their dominant sense 97.63% of the time for Expansion, 94.72% for Contingency, 93.43% for Comparison and 84.10% for Temporal. Temporal is the most ambiguous class. Problem cases include since (52.17% Contingency) and while (66.07% Comparison).
  • One-vs-rest classification: Comparison versus all other relations scores 91.28% on all data and 97.23% on explicit relations only. Every binary task on explicit relations reaches about 94% or higher.
  • Strongest pair: The strongest association between neighboring relations is explicit Comparison followed by implicit Contingency (χ² = 20.1, p β‰ˆ 0.000007).

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

Fields: Linguistics & languages

Topics: Discourse analysis

Cited by

76 papers of our library cite it (of the 140673 whose bibliographies we read); 46 of them have a page here, 30 not yet.

Cite

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

@inproceedings{pitler2008easily,
  title = {{Easily Identifiable Discourse Relations}},
  author = {Pitler, Emily and Raghupathy, Mridhula and Mehta, Hena and Nenkova, Ani and Lee, Alan and Joshi, Aravind},
  booktitle = {Coling 2008: Companion volume: Posters},
  year = {2008},
  url = {https://aclanthology.org/C08-2022/}
}

References

10 references, 4 of them with a card of ours. β†— leads to the work, Β«findΒ» searches for one we have no id for yet.

  1. Hobbs, J. 1979. Coherence and coreference. Cognitive Science, 3:67–90. find
  2. Knott, A. and T. Sanders. 1998. The classification of coherence relations and their linguistic markers: An exploration of two languages. Journal of Pragmatics, 30(2):135–175. find
  3. Mann, W. and S. Thompson. 1988. Rhetorical structure theory: Toward a functional theory of text organization. Text, 8:243–281. find our card
  4. Marcus, M.P., B. Santorini, and M.A. Marcinkiewicz. 1994. Building a Large Annotated Corpus of English: The Penn Treebank. Computational Linguistics, 19(2):313–330. find our card
  5. McKeown, Kathleen R. 1985. Text generation: using discourse strategies and focus constraints to generate natural language text. Cambridge University Press, New York, NY, USA. find
  6. Miltsakaki, E., N. Dinesh, R. Prasad, A. Joshi, and B. Webber. 2005. Experiments on sense annotations and sense disambiguation of discourse connectives. In Proceedings of the Fourth Workshop on Treebanks and Linguistic Theories (TLT2005). find
  7. Miltsakaki, Eleni, Livio Robaldo, Alan Lee, and Aravind Joshi. 2008. Sense annotation in the penn discourse treebank. Computational Linguistics and Intelligent Text Processing, Lecture Notes in Computer Science, 4919:275–286. find
  8. Moore, J. and P. Wiemer-Hastings. 2003. Discourse in computational linguistics and artificial intelligence. find
  9. Prasad, R., N. Dinesh, A. Lee, E. Miltsakaki, L. Robaldo, A. Joshi, and B. Webber. 2008. The penn discourse treebank 2.0. In Proceedings of the 6th International Conference on Language Resources and Evaluation (LREC). find our card
  10. Wellner, B., J. Pustejovsky, C. Havasi, R. Sauri, and A. Rumshisky. 2006. Classification of discourse coherence relations: An exploratory study using multiple knowledge sources. In Proceedings of the 7th SIGDIAL Workshop on Discourse and Dialogue. find our card

↑ back to top