Elamur Research⌕
☰
home / papers / 2024 / Fusing ISO 24617-2 Dialogue Acts and Application-Specific Semantic Content Annotations

Fusing ISO 24617-2 Dialogue Acts and Application-Specific Semantic Content Annotations

Paper · 2024 · Andrei Malchanau, Volha Petukhova and Harry Bunt

Summary

The paper addresses a gap in dialogue act annotation. Most schemes, including the ISO 24617-2 standard (DiAML), describe the communicative function of an utterance in depth but leave its semantic content open or unspecified. Real dialogue systems need both parts. The authors therefore propose a multi-scheme design. The domain-independent DiAML layer holds the functional side of each dialogue act. Separate, application-defined XSD schemes supply the semantic content, using the plug-in mechanism introduced in the 2019 second edition of the standard.

They also describe an XSD architecture that mixes namespace design patterns. With it, combined annotations can be formally validated with standard tools, and code bindings for system modules can be generated automatically. The approach is shown on three use cases of increasing semantic complexity:

  • intent and slot filling,
  • term-based information retrieval,
  • rich situation and experience modelling.

The value is practical. Existing annotations can be reused, checked and extended across applications, and annotated corpora connect directly to running dialogue software.

Key points

  • In DiAML, a dialogue act entity structure has seven parts: sender, addressees, other participants, dimension, communicative function, dependence relations, and qualifiers. Link structures relate acts through rhetorical relations.
  • Core types live in DiAML_Types.xsd under the diaml namespace. Namespace-less containers (corpus, sessions, segments, inter-module messages) live in DiAML_Containers.xsd. The containers take on the application namespace (the chameleon pattern), and content schemes such as SemAF parts are imported in the heterogeneous style.
  • Applications redefine <dialogueAct> in their own namespace (examples include DBOX and LICA) and add an optional semanticContent element. This avoids forward references while keeping validation possible.
  • Validation is done with Oxygen, and Java classes are generated automatically via JAXB.
  • The LICA health-dialogue example annotates a doctor's claim and supporting evidence with these attributes:
    • stance, expertise, importance, framing, and persuasion strategy;
    • justification links that tie each piece of evidence to the claim.

Interesting ideas

  • I1. Keep a small, general "function" layer separate from a swappable, domain-specific "content" layer, so that one core scheme serves many applications without bloating.
  • I2. Define annotation formats as formal schemas from which both validators and data-binding code are generated, so the corpus format and the system's internal types cannot drift apart.
  • I3. Leave semantic content optional for dialogue-control acts, since their meaning lies mostly in function and dimension, and spend annotation effort only where content matters.
  • I4. Represent argumentation as claim and evidence units joined by justification relations, each carrying stance, expertise, framing and strategy attributes.
  • I5. Grade the design against use cases of increasing semantic richness (slots, then terms, then situations) to test whether one architecture scales.

For Mova

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

  • I1 (frames): Speech-act labels in prag could form a fixed core, with frame/VerbNet content attached as a pluggable payload per task domain.
  • I2 (learn): Generating Rust types and validators for silver-data records from one schema would catch malformed witness-parser outputs before training.
  • I3 (frames): prag could skip building full frames for feedback or turn-management utterances, saving work and avoiding spurious roles.
  • I4 (world): Fable morals or persuasive texts could be stored as claim–evidence graphs, so answers can cite which statement supports which conclusion.
  • I5 (answer): Evaluate readers on a ladder from slot-like factual questions up to full situation-model questions, to see where explanations break down.

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, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Dialogue systems · 624 worksSpeech act theory · 229 works

Cite

Cite as Andrei Malchanau, Volha Petukhova and Harry Bunt. 2024. Fusing ISO 24617-2 Dialogue Acts and Application-Specific Semantic Content Annotations. Proceedings of the 20th Joint ACL - ISO Workshop on Interoperable Semantic Annotation @ LREC-COLING 2024.

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

@inproceedings{malchanau2024fusing,
  title = {{Fusing ISO 24617-2 Dialogue Acts and Application-Specific Semantic Content Annotations}},
  author = {Andrei Malchanau, Volha Petukhova and Harry Bunt},
  booktitle = {Proceedings of the 20th Joint ACL - ISO Workshop on Interoperable Semantic Annotation @ LREC-COLING 2024},
  year = {2024},
  doi = {10.63317/3y82n7g7rb8b},
  url = {https://doi.org/10.63317/3y82n7g7rb8b}
}

References

37 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Laura Banarescu et al. 2013. Abstract meaning representation for Sembanking. In Proceedings of the 7th Linguistic Annotation Workshop and Interoperability with Discourse.
  2. Tim Benson. 2012. Principles of health interoperability HL7 and SNOMED. Springer Science & Business Media. ↗
  3. Abhinav Bhatt. 2022. Personalized medical arguments generation. Master’s thesis, Saarland University. find
  4. Olivier Bodenreider. 2004. The Unified Medical Language System (umls): integrating biomedical terminology. Nucleic acids research, 32(suppl_1):D267–D270. ↗
  5. Claire Bonial et al. 2020. Dialogue-AMR: abstract meaning representation for dialogue. In Proceedings of the 12th Language Resources and Evaluation Conference. https://dialogbank.lsv.uni-saarland.de/ 13 https://catalog.ldc.upenn.edu/LDC2017S11 ↗
  6. Harry Bunt. 2014. Annotations that effectively contribute to semantic interpretation. In Computing Meaning: Volume 4, pages 49–69. Springer. ↗
  7. Harry Bunt. 2015. On the principles of semantic annotation. In Proceedings of the 11th Workshop on Interoperable Semantic Annotation (ISA-11). find
  8. Harry Bunt. 2019. Plug-ins for content annotation of dialogue acts. In 15th Workshop on Interoperable Semantic Annotation (ISA-15). find
  9. Harry Bunt. 2024. Combining semantic annotation schemes through interlinking. In 20th Workshop on Interoperable Semantic Annotation (ISA-20). find
  10. Ziming Cheng, Zuchao Li, and Hai Zhao. 2022. Bibl: AMR parsing and generation with bidirectional bayesian learning. In 29th International Conference on Computational Linguistics. find
  11. R.L. Costello. 2006. Ml schemas: Best practices. multi-schema project: Zero, one, or many namespaces? find
  12. Susan T Fiske. 2018. Stereotype content: Warmth and competence endure. Current directions in psychological science, 27(2):67–73. ↗
  13. Maarten Grootendorst. 2020. Keybert: Minimal keyword extraction with bert. find
  14. Bushra S Guenoun and Julian J Zlatev. 2023. Sending signals: Strategic displays of warmth and competence. Working Paper 23-051. find
  15. ISO. 2012a. Language resource management – Semantic annotation framework – Part 2: Dialogue acts. ISO 24617-2. ISO Central Secretariat, Geneva. find
  16. ISO. 2012b. Language Resource Management-Semantic Annotation Framework (SemAF)- Part 1: Time and events. ISO 24617-1. ISO Central Secretariat, Geneva. find
  17. ISO. 2014. Language Resource Management- Semantic Annotation Framework (Se mAF)- Part 4: Semantic roles ISO 24617-4. ISO Central Secretariat, Geneva. find
  18. ISO. 2016. Language resource management - Semantic annotation framework (SemAF) Part 8, Semantic relations in discourse ISO 24617-8. ISO Central Secretariat, Geneva. find
  19. ISO. 2019a. Language resource management - Semantic annotation framework- Part 7: Spatial information (ISO-Space). ISO 24617-7. ISO Central Secretariat, Geneva. find
  20. ISO. 2019b. Language Resource Management-Semantic Annotation Framework (SemAF)- Part 9: Reference annotation framework (RAF) ISO 24617-9. ISO Central Secretariat, Geneva. find
  21. ISO. 2019c. Language Resource Management: Semantic Annotation Frame work (SemAF)-Part 12: Quantification. ISO 24617-12. ISO Central Secretariat, Geneva. find
  22. Hye-Kyeong Ko and Minho Yang. 2017. An effective xml schema conversion technique for improving xml document reusability using pattern list. International Journal of Internet, Broadcasting and Communication, 9(2):11–19. find
  23. Valeria Lapina and Volha Petukhova. 2017. Classification of modal meaning in negotiation dialogues. In Proceedings of the 13th Joint ISO-ACL Workshop on Interoperable Semantic Annotation (ISA-13). ↗
  24. Stefan Larson and Kevin Leach. 2022. A survey of intent classification and slot-filling datasets for task-oriented dialog. ↗
  25. Andrei Malchanau, Volha Petukhova, and Harry Bunt. 2018. Towards integration of cognitive models in dialogue management: designing the virtual negotiation coach application. Dialogue & Discourse, 9(2):35–79. ↗
  26. Bonan Min and Ralph Grishman. 2012. Challenges in the knowledge base population slot filling task. In LREC, pages 1137–1142. find
  27. Martha Palmer, Daniel Gildea, and Paul Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. Computational linguistics, 31(1):71–106. ↗
  28. Volha Petukhova et al. 2018. Understanding questions and extracting answers: Interactive quiz game application design. In Human Language Technology. Challenges for Computer Science and Linguistics, pages 246–261. Springer. find
  29. James Pustejovsky and Nancy Ide, editors. 2017. Handbook of Linguistic Annotation. Springer. find
  30. Janet G Schnall and Susan Fowler. 2013. Medlineplus. gov: quality health information for your patients. AJN The American Journal of Nursing, 113(9):64–65. find
  31. Christopher Stevens, Harmen de Weerd, Fokie Cnossen, and Niels Taatgen. 2016. A metacognitive agent for training negotiation skills. In Proceedings of the 14th International Conference on Cognitive Modeling (ICCM 2016). find
  32. Bahar Tarakameh. 2019. Integration of business process models into dialogue management: patient self-assessment use case. Master’s thesis, Saarland University. find
  33. Pavlo Vasylenko et al. 2023. Incorporating graph information in transformer-based amr parsing. In Findings of the Association for Computational Linguistics: ACL 2023, page 1995–2011. ↗
  34. Clara Wan Ching Ho and Volha Petukhova. 2024. Towards generation of personalised health intervention messages. In Proceedings of the 1st Workshop on Patient-oriented Language Processing (CL4HEALTH), LREC-COLING 2024). find
  35. Martin Wolf et al. 2019. Term-based extraction of medical information: Pre-operative patient education use case. In Conference on Recent Advances in Natural Language Processing. ↗
  36. Qing Zeng et al. 2007. Term identification methods for consumer health vocabulary development. Journal of medical Internet research, 9 (1). ↗
  37. Jiawei Zhou et al. 2021. Structure-aware fine-tuning of sequence-to-sequence transformers for transition-based AMR parsing. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/malchanau-2024-fusing-iso-24617-2-dialogue/