Elamur Research⌕
☰
home / papers / 2014 / A Model for Processing Illocutionary Structures and Argumentation in Debates

A Model for Processing Illocutionary Structures and Argumentation in Debates

Paper · 2014 · Kasia Budzynska, Mathilde Janier, Chris Reed et al.

Summary

The paper tackles automatic recognition of the illocutionary structure that underlies argumentation in spoken debate, where speakers often argue indirectly, for example by asking questions rather than asserting. It builds on Inference Anchoring Theory (IAT), which ties three things together: dialogue units, the illocutionary forces that link units to their propositional content, and the inferences or conflicts between those contents. In IAT, transitions between possibly non-adjacent moves serve as anchors for argumentative relations. The development data is MM2012, a transcript corpus of the BBC Radio 4 debate show The Moral Maze. The authors propose a small taxonomy of question-like forces, including polymorphic types that deliberately combine two forces. They then implement recognition as hand-written Dislog rules driven by lexical and morphological markers. On manually annotated texts, the forces of correctly segmented units were assigned acceptably in 78% of cases once polymorphic labels are allowed. The work matters because it lets questions that carry hidden assertions enter an argument graph, which a plain dialogue-act tagger cannot do.

Key points

  • MM2012 contains about 65,000 words of The Moral Maze transcripts, a debate format the authors describe as agreement-seeking.
  • IAT is compared with RST, SDRT, KoS and the SwDA tagset (over 200 tags, 60 basic). In the authors' view, none of these captures how pairs of locutions establish inference.
  • The scheme defines 12 simple or polymorphic types for questions, such as RQ, AQ, PQ, ACh, PCh, PQ-AQ and RQ-PCh. These are recognised with 18 Dislog rules over cue markers like "why should", "was it" and "would you".
  • About 63% of the markers point to a single force, yet just around 62% of units in the text can be considered unambiguous before any context is taken into account. The most frequent ambiguities are PQ-AQ (13%) and AQ-RQ (10%).
  • Over 153 correctly segmented units, 38% were assigned a correct basic type, 40% a correct polymorphic type and 22% were wrong. About 65% of the polymorphic cases mix different strengths of assertion.

Interesting ideas

  • I1. When cues cannot separate two forces, the system emits an explicit combined label such as RQ-AQ instead of forcing a choice, which keeps the ambiguity visible for later stages to resolve.
  • I2. A question can carry two functions at once: it asserts a belief, and it seeks agreement while lowering the speaker's commitment ("I was only asking").
  • I3. Argumentative relations are anchored in transitions between moves, which may be non-adjacent, rather than in rhetorical relations between sentences. This separates "who responded to what" from the question of "what follows from what".
  • I4. A small rule set over surface markers was chosen so that interactions between competing rules stay easy to manage and inspect.
  • I5. Evaluation is split into correct basic, correct polymorphic and wrong assignments, rather than reported as a single accuracy figure.

For Mova

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

  • I1 (frames): prag could output paired speech-act labels like question-or-assertion with the cue that triggered each, letting the world stage resolve them from context.
  • I2 (frames): speech-act frames could record both an asserted proposition and a weaker commitment level, so that interrogative claims still feed arguments and implicature.
  • I3 (world): the world of a text could keep a move-transition layer, so that a later answer stage can explain which exchange established an inference or a conflict.
  • I4 (frames): cue-pattern rules for speech acts can be kept few and traceable, each firing rule logged as the explanation for the label it assigned.
  • I5 (learn): speech-act evaluations could report how often an underspecified label was right, which shows whether ambiguity is being handled honestly or just hidden.

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

🔬Centre National de la Recherche ScientifiqueParis · France104 works · 60,877 citations🎓University of PotsdamPotsdam · Germany36 works · 5,236 citations🔬Institut de Recherche en Informatique de ToulouseToulouse · France23 works · 1,667 citations🎓University of DundeeDundee · United Kingdom2 works · 643 citations🏛️Polish Academy of SciencesWarsaw · Poland8 works · 512 citations

Also on the byline: Institute of Philosophy and Sociology.

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksPhilosophy · 211 works

🧭 Topics 1Research subjects and languages the work is about.Speech act theory · 160 works

Cite

Cite as Kasia Budzynska, Mathilde Janier, Chris Reed, Patrick Saint-Dizier, Manfred Stede and Olena Yakorska. 2014. A Model for Processing Illocutionary Structures and Argumentation in Debates. Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14).

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

@inproceedings{budzynska2014model,
  title = {{A Model for Processing Illocutionary Structures and Argumentation in Debates}},
  author = {Kasia Budzynska, Mathilde Janier, Chris Reed, Patrick Saint-Dizier, Manfred Stede and Olena Yakorska},
  booktitle = {Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14)},
  year = {2014},
  doi = {10.63317/52zr77iewo78},
  url = {https://doi.org/10.63317/52zr77iewo78}
}

References

24 references, 5 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. (2003). Logics of Conversations. Cambridge University Press. find
  2. Budzynska, K. and Reed, C. (2011). Whence inference? In University of Dundee Technical Report. find
  3. Budzynska, K., Janier, M., Reed, C., and Saint-Dizier, P. (2013). Theoretical foundations for illocutionary structure parsing. In Proc. CMNA. find
  4. Carletta, J. (1997). The reliability of a dialogue structure coding scheme. Computational Linguistics, 23(1):13– 32. ↗
  5. Fiedler, Armin and Horacek, Helmut. (2007). Argumentation within deductive reasoning. International Journal of Intelligent Systems, 22(1):49–70. ↗
  6. Ginzburg, J., (2012). The Semantics of Dialogue. Cambridge University Press. find
  7. Grosz, B. and Sidner, C. (1986). Attention, intention and the structure of discourse. Computational Linguistics, 12(3):175–204. find
  8. Hernault, H., Prendinger, H., du Verle, D., and Ishizuka, M. (2010). Hilda: A discourse parser using support vector machine classification discourse and dialogue. Discourse and Dialogue, 1(3). ↗
  9. Mackenzie, J. (1990). Four dialogue systems. Studia Logica, 51:567–583. ↗
  10. Mann, W. C. and Thompson, S. (1988). Rhetorical structure theory: Toward a functional theory of text organization. Text, 8(3):243–281. ↗
  11. Mann, W. C. (2002). Dialogue macrogame theory. http://www-bcf.usc.edu/ billmann/dialogue/SIG-DMT-PF6.htm. ↗
  12. Marcu, D. (1997). The rhetorical parsing of natural language texts. ACL. find
  13. Marcu, D. (2000). The Theory and Practice of Discourse Parsing and Summarization. MIT Press. ↗
  14. Moeschler, J., (2002). Speech act theory and the analysis of conversation, pages 239–261. John Benjamins. Peldszus, Andreas and Stede, Manfred. (to appear). From argument diagrams to argumentation mining in texts: a survey. International Journal of Cognitive Informatics and Natural Intelligence (IJCINI). find
  15. Reed, C., Palau, R. Mochales, Rowe, G., and Moens, M.-F. (2008). Language resources for studying argument. In Proc. of LREC 2008. find
  16. Reed, C., Wells, S., Budzynska, K., and Devereux, J. (2010). Building arguments with argumentation: the role of illocutionary force in computational models of argument. Frontiers in Artificial Intelligence and Applications, 216:415–426. ↗
  17. Reed, C. (1998). Generating arguments in natural language. Unpublished doctoral dissertation, University College London. find
  18. Saint-Dizier, P. (2012). Processing natural language arguments with the <textcoop> platform. Journal of Argument and Computation, 3(1):49–82. find
  19. Searle, J. and Vanderveken, D. (1985). Foundations of Illocutionary Logic. Cambridge University Press. find
  20. Searle, J. (1969). Speech Acts: An essay in the philosophy of language. Cambridge. ↗
  21. Stede, M. (2012). Discourse Processing. Morgan and Claypool Publishers. find
  22. Stolcke, A., Ries, K., Coccaro, N., Shriberg, E., Bates, R., Jurafsky, D., Taylor, P., Martin, R., Meteer, M., and Ess-Dykema, C. Van. (2000). Dialogue act modeling for automatic tagging and recognition of conversational speech. Computational Linguistics, 26(3):339–371. ↗
  23. Walton, D., Reed, C., and Macagno, F. (2008). Argumentation Schemes. Cambridge University Press. find
  24. Winterstein, G. (2012). What but-sentences argue for: An argumentative analysis of but. Lingua, 122:1864–1885. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/budzynska-2014-model-processing-illocutionary-structures-argumentation/