Elamur Research⌕
☰
home / papers / 2012 / ISO 24617-2: A semantically-based standard for dialogue annotation

ISO 24617-2: A semantically-based standard for dialogue annotation

Summary

This paper presents the final, ISO-approved version of ISO 24617-2, an application-independent standard for annotating dialogue acts in typed, spoken and multimodal interaction. It explains what changed since the 2010 draft (ISO DIS 24617-2:2010).

The standard treats dialogue acts as operations that update participants' information states. It defines annotations through the Dialogue Act Markup Language DiAML, which has three parts: an abstract syntax, a formal semantics and a concrete XML syntax.

The final version makes four main changes:

  • It adds rhetorical relations between dialogue acts.
  • It gives DiAML a full compositional semantics. Working this out showed that the draft's way of encoding functional and feedback dependences as separate links was semantically wrong, so these dependences are now attributes of the dialogue act itself.
  • It derives the XML reference format systematically from the abstract syntax, using the principle of an "ideal concrete syntax".
  • It reports an incremental, token-based recognizer trained on AMI data with high F-scores, showing that the scheme also works for automatic annotation.

Key points

  • The scheme defines nine semantic dimensions, including Task, Auto-Feedback, Allo-Feedback, Turn Management and Social Obligation Management. These support multidimensional tagging and segmentation, with functional segments as the annotation unit.
  • Rhetorical relations are encoded as rhetoricalLink elements with an @rhetoRel attribute. The standard deliberately leaves the inventory of relations open for each project to choose.
  • Function qualifiers can express conditionality, certainty and sentiment. Their semantics either narrows the update a function performs or adds an extra update.
  • The automatic recognizer used 64 local classifiers trained on AMI, plus global classifiers that take the local outputs over a window of ±5 tokens.
    • With Bayes Net and Ripper, F-scores ranged from about 70 (Partner Communication Management) to 98.6 (Social Obligation Management).
  • The ANVIL annotation tool was extended to export ISO 24617-2 annotations in DiAML format.

Interesting ideas

  • I1. Separating content from format: an annotation scheme can be defined as an abstract syntax plus a formal semantics, with XML as only one possible rendering. Any two formats that faithfully render the same structures can then be converted into each other without losing meaning. This is useful for any project that has to support several file formats.
  • I2. Ideal concrete syntax: the serialization format should be complete (every abstract structure can be written down) and unambiguous (each written form maps to exactly one structure). Building XML elements directly from the abstract model gives transparent formats and is a good checklist for designing new annotation schemas.
  • I3. Checking the schema against its semantics: writing a formal semantics exposed a modelling error in the draft. Dependences had been encoded as link structures, but they actually belong inside the dialogue acts. Testing a draft scheme against an explicit semantics in this way can catch design errors before a standard is finalized.
  • I4. Leaving contested inventories open: the standard fixes the slot for rhetorical relations but not the list of relation labels. This keeps the standard stable while allowing project-specific label sets.
  • I5. Stacking local and global classifiers incrementally: narrow per-label classifiers feed a higher-level classifier that also sees predictions for neighbouring tokens. This is a simple way to do online multi-label tagging of streaming text or speech.

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: Speech act theory

Cited by

64 papers of our library cite it (of the 140673 whose bibliographies we read); 7 of them have a page here, 57 not yet.

Cite

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

@inproceedings{bunt2012iso,
  title = {{ISO 24617-2: A semantically-based standard for dialogue annotation}},
  author = {Bunt, Harry and Alexandersson, Jan and Choe, Jae-Woong and Fang, Alex Chengyu and Hasida, Koiti and Petukhova, Volha and Popescu-Belis, Andrei and Traum, David},
  booktitle = {Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12)},
  year = {2012},
  doi = {10.63317/2vxg5i9ctqt8},
  url = {https://doi.org/10.63317/2vxg5i9ctqt8}
}

References

22 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ahn, R. (2001) Agents, Object and Events: A computational approach to knowledge, observation and communication. PhD Thesis, Eindhoven University of Technology. find
  2. Allen J, Core M (1997) DAMSL: Dialogue Act Markup in Several Layers (Draft 2.1). Technical Report. University of Rochester, Rochester, NY. find
  3. Asher, N., and A. Lascarides (2003) Logics of Conversation. Cambridge University Press, UK. find
  4. Bunt, H. (1994) Context and Dialogue Control. Think Quarterly 3, 19–34. find
  5. Bunt, H. (2000a) Dialogue Pragmatics and Context Specification. In: Harry Bunt & William Black (eds.) Abduction, Belief and Context in Dialogue. Benjamins, Amsterdam, p. 81–150. ↗
  6. Bunt, H. (2009a) The DIT ++ taxonomy for functional dialogue markup. In D. Heylen, C. Pelachaud, R. Catizone and D. Traum (eds.) Proceedings of AAMAS 2009 Workshop “Towards a Standard Markup Language for Embodied Dialogue Acts’, Budapest, pp. 13-24. find
  7. Bunt, H. (2010). Multifunctionality in Dialogue. Computers, Speech, and Language 22, 224-245. ↗
  8. Bunt, H. (2011). The semantics of dialogue acts. In Proceedings of the 9 th International Conference on Computational Semantics (IWCS 2011), Oxford, pp. 1-14. find
  9. Bunt, H., J. Alexandersson, J. Carletta, J.-W. Chae, A. Fang, K. Hasida, K. Lee, V. Petukhova, A. Popescu-Belis, L. Romary, C. Soria, and D. Traum (2010). Towards an ISO standard for dialogue act annotation. In Proceedings of the 7 th International Conference on Language Resources and Systems (LREC 2010), Malta, pp. 2548–2558. find our card
  10. Bunt, H., M. Kipp, and V. Petukhova (2012). Towards an ISO standard for dialogue act annotation. In Proceedings of the 8 th International Conference on Language Resources and Systems (LREC 2012), Istanbul. find our card
  11. Cooper, R. (2000) Information states, attitudes and dependent record types. In: Cavedon L, Blackburn P, Braisby N, Shimojima A (eds) Logic, Language and Computation, Vol 3, CSLI Publications, Stanford, pp 85–106. find
  12. Hobbs, J. (1985) On the Coherence and Structure of Discourse. CSLI Research Report 85-37. find
  13. Hovy E., and E. Maier (1995) Parsimonious or profligate: how many and which discourse structure relations? Unpublished manuscript. find
  14. ISO DIS 24617-2 (2010) Language resource management – Semantic annotation framework (SemAF), Part 2: Dialogue acts. ISO, Geneva. find
  15. Kipp, M. (2001) ANVIL - a Generic Annotation Tool for Multimodal Dialogue. In Proceedings of Eurospeech 2001, pp. 1367–1370. find
  16. Kipp, M. (2008) Spatiotemporal coding in ANVIL. In Proceedings of LREC 2008, Marrakech. find
  17. Mann, W. and Thompson, S. (1988) Rhetorical structure theory: toward a functional theory of text organisation. MIT Press, Cambridge, MA. find
  18. Petukhova, V. and H. Bunt (2011) Incremental dialogue act understanding. In Proceedings of the 9 th International Conference on Computational Semantics (IWCS 2011), Oxford, pp. 235-244. find
  19. Poesio, M. and D. Traum (1997) Conversational actions and discourse situations. Computational Intelligence 13(3) pp. 309 – 347 ↗
  20. 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 LREC 2008, Marrakech. find our card
  21. Tonelli, S., G. Riccardi, R. Prasad and A. Joshi (2009) Annotation of Discourse Relations for Conversational Spoken Dialogs. In Proceedings LREC 2010, Malta. find our card
  22. Traum, D. & S. Larsson (2003) The Information State Approach to Dialogue Act Management. In J. van Kuppevelt & R. Smith (eds.) Current and new Directions in Discourse and Dialogue, Kluwer, Dordrecht. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/bunt-2012-iso-24617-2-semantically-based/