Elamur Research⌕
☰
home / papers / 2018 / The ADELE Corpus of Dyadic Social Text Conversations:Dialog Act Annotation with ISO 24617-2

The ADELE Corpus of Dyadic Social Text Conversations:Dialog Act Annotation with ISO 24617-2

Paper · 2018 · Emer Gilmartin, Christian Saam, Brendan Spillane et al.

Summary

Social, chatty dialog has received far less descriptive attention than goal-driven dialog, yet companion-style chatbots and voice assistants need data about how it works. The authors review accounts of casual conversation, including Ventola's phase model of greeting, approach, centring and leave-taking. They also examine how existing dialog act schemes, especially the ISO 24617-2 standard, cover non-task talk.

They then built the ADELE Corpus: 193 two-person text chats collected through a "getting to know you" game. Each participant received a fictitious persona and had to uncover the partner's traits and any shared ones. The dialogs were annotated with a modified ISO 24617-2 scheme. The modifications add topic tags and separate first mentions of information from follow-up comments, and the authors propose extra acts for greeting and leave-taking sequences. Preliminary analyses cover conversation length, utterance length and lexical density, and relate the text chats to spoken language.

The resource matters because informal text chat appears closer to speech than to formal writing. It gives social dialog systems annotated training data for the onboarding phase of a conversation.

Key points

  • Data: 193 dyadic text dialogs from 37 participants (26 male, 11 female, aged 18–43), collected over two months in late 2016 at Trinity College Dublin. Participants were native English speakers or had IELTS 6.5 or higher.
  • Collection platform:
    • a Chrome extension for Slack served as the Dialog Interface;
    • a Java JAX-RS/Jersey RESTful "Matching Engine" paired users and assigned random personas;
    • PostgreSQL stored the data.
  • Incentive design: players earned 1 point per utterance, 1 per discovered partner trait, and 5 per discovered commonality. This was meant to push them toward friendly chat rather than interrogation.
  • Coverage of social acts in existing schemes:
    • The ISO standard's Social Obligations Management dimension has only nine social functions, such as initialGreeting, apology, thanking and initialGoodbye.
    • Petukhova's survey of 14 schemes found greeting functions in 10, introductions in 4, goodbyes in 6, apologies in 5 and thanking in 5.
  • Annotation method:
    • Annotation followed ISO-annotated Switchboard examples from the Dialog Bank, after a two-annotator pilot.
    • Inform acts carry [topic] tags.
    • Any repeat mention of known information is relabelled as a comment.
    • New greeting and leave-taking acts are proposed.

Interesting ideas

  • I1. An inform act that conveys information already given is relabelled as a comment, which separates task-advancing turns from social follow-up talk.
  • I2. Dialog act labels are paired with a lexical [topic] tag, so each turn records both what the speaker does and which subject it concerns.
  • I3. The scoring rules shape behaviour during elicitation: rewarding turns and shared facts produces natural chat instead of question-answer interviews.
  • I4. Conversation is modelled as phases (greeting, approach, centring, leave-taking), with approach and centring able to recur and alternate in the middle.
  • I5. Persona-driven dialogs with known ground-truth attributes make it possible to check exactly which facts each participant should have learned.

For Mova

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

  • I1 (frames): The speech-act layer in prag could mark whether an assertion introduces a new fact or merely reacts to a known one, which helps decide what actually updates the text's world.
  • I2 (world): Pairing each act with the entity or attribute it touches would let the world model track which facts about a character were revealed and when.
  • I3 (learn): In self-play or synthetic data generation, reward structures could be tuned so the generated dialogs show realistic conversational behaviour rather than degenerate interrogation.
  • I4 (answer): Readers answering questions about dialogs or stories could use a coarse phase structure to locate where introductions or closings happen and interpret turns accordingly.
  • I5 (learn): Synthetic dialogs generated from hidden personas would give an evaluation set where Mova's extracted world facts can be scored against a known answer key.

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

🏢IBMArmonk · United States85 works · 66,118 citations🎓Trinity College DublinDublin · Ireland17 works · 1,222 citations

Also on the byline: Accenture (Norway).

📚 Shelves 4The 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 works

🧭 Topics 2Research subjects and languages the work is about.Dialogue systems · 401 worksSpeech act theory · 160 works

Cite

Cite as Emer Gilmartin, Christian Saam, Brendan Spillane, Maria O'Reilly, Ketong Su, Arturo Calvo, Loredana Cerrato, Killian Levacher, Nick Campbell and Vincent Wade. 2018. The ADELE Corpus of Dyadic Social Text Conversations:Dialog Act Annotation with ISO 24617-2. Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018).

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

@inproceedings{gilmartin2018adele,
  title = {{The ADELE Corpus of Dyadic Social Text Conversations:Dialog Act Annotation with ISO 24617-2}},
  author = {Emer Gilmartin, Christian Saam, Brendan Spillane, Maria O'Reilly, Ketong Su, Arturo Calvo, Loredana Cerrato, Killian Levacher, Nick Campbell and Vincent Wade},
  booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)},
  year = {2018},
  doi = {10.63317/2gi8fex93b55},
  url = {https://doi.org/10.63317/2gi8fex93b55}
}

References

26 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ai, H. and Lu, X. (2010). A web-based system for automatic measurement of lexical complexity. In 27th Annual Symposium of the Computer-Assisted Language Instruction Consortium. Amherst, MA. ↗
  2. Allen, J., Ferguson, G., and Stent, A. (2001). An Architecture for More Realistic Conversational Systems. In International Conference on Intelligent User Interfaces: Proceedings of the 6th International Conference on Intelligent User Interfaces, volume 14, pages 1–8. find
  3. Aubrey, A. J., Marshall, D., Rosin, P. L., Vandeventer, J., Cunningham, D. W., and Wallraven, C. (2013). Cardiff conversation database (ccdb): A database of natural dyadic conversations. In Computer Vision and Pattern Recognition Workshops (CVPRW), 2013 IEEE Conference on, pages 277–282. IEEE. find
  4. Biber, D., Johansson, S., Leech, G., Conrad, S., Finegan, E., and Quirk, R. (1999). Longman Grammar of Spoken and Written English, volume 2. Longman London. find
  5. Brown, G. and Yule, G. (1983). Teaching the Spoken Language, volume 2. Cambridge University Press. find
  6. Bunt, H., Petukhova, V., Malchanau, A., Wijnhoven, K., and Fang, A. C. (2016). The dialogbank. In LREC.
  7. Bunt, H. (2006). Dimensions in Dialogue Act Annotation. In Proc. of LREC, volume 6, pages 919–924. find
  8. Core, M. G. and Allen, J. (1997). Coding Dialogs with the DAMSL Annotation Scheme. In AAAI Fall Symposium on Communicative Action in Humans and Machines, pages 28–35. Boston, MA. find
  9. Eggins, S. and Slade, D. (2004). Analysing Casual Conversation. Equinox Publishing Ltd. find
  10. Fairclough, N. (1992). Discourse and social change. Polity, Cambridge. find
  11. Gilmartin, E., Spillane, B., Saam, C., O’Reilly, M., Su, K., Cowan, B. R., Levacher, K., Calvo Dehesa, A., Cerrato, L., Campbell, N., and Wade, V. (2017). Annotation of Greeting, Introduction, and Leavetaking in Text Dialogues Using ISO 24617-2. In Proceedings of the Thirteenth Joint ACL - ISO Workshop on Interoperable Semantic Annotation, Montpellier, France. find
  12. Grice, H. P. (1975). Logic and conversation. In John P. Kimball, et al., editors, Syntax and semantics. Vol.3, Speech acts. Academic Press, New York [etc.] ; London. ↗
  13. Halliday, M. A. (1989). Spoken and written language. find
  14. ISO. (2012). ISO 24617-2:2012 - Language resource management – Semantic annotation framework (SemAF) – Part 2: Dialogue acts. International Organization for Standardization, Geneva, Switzerland. find
  15. Jakobson, R. (1960). Linguistics and poetics. In Th. A. Sebeok, editor, Style in language, pages 350–377. MA: MIT Press, Cambridge. find
  16. Jurafsky, D., Van Ess-dykema, C., et al. (1997). Switchboard Discourse Language Modeling Project (Final Report). find
  17. Kim, S. N., Cavedon, L., and Baldwin, T. (2010). Classifying Dialogue Acts in One-on-One Live Chats. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 862–871. Association for Computational Linguistics.
  18. Laver, J. (1975). Communicative Functions of Phatic Communion. In Adam Kendon, et al., editors, Organization of behavior in face-to-face interaction, pages 215– 238. Mouton, Oxford, England. ↗
  19. Malinowski, B. (1936). The Problem of Meaning in Primitive Languages. In The meaning of meaning: a study of the influence of language upon thought and of the science of symbolism, pages 296–336. Kegan Paul, Trench, Trübner, London, 4th ed. rev edition. find
  20. McCowan, I., Carletta, J., Kraaij, W., Ashby, S., Bourban, S., Flynn, M., Guillemot, M., Hain, T., Kadlec, J., and Karaiskos, V. (2005). The AMI Meeting Corpus. In Proceedings of the 5th International Conference on Methods and Techniques in Behavioral Research, volume 88. find
  21. Petukhova, V. (2011). Multidimensional Dialogue Modelling. Ph.D. thesis, Tilburg University, Tilburg, Netherlands. find
  22. Schneider, K. P. (1988). Small Talk: Analysing Phatic Discourse, volume 1. Hitzeroth Marburg. find
  23. Shriberg, E., Dhillon, R., Bhagat, S., Ang, J., and Carvey, H. (2004). The ICSI Meeting Recorder Dialog Act (MRDA) Corpus. Technical report, International Computer Science Institute, Berkeley. ↗
  24. Traum, D. (1999). Speech Acts for Dialogue Agents. Foundations of Rational Agency, 14:169–202. ↗
  25. Ure, J. (1971). Lexical density and register differentiation. Applications of linguistics, pages 443–452. find
  26. Ventola, E. (1979). The Structure of Casual Conversation in English. Journal of Pragmatics, 3(3):267–298. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/gilmartin-2018-adele-corpus-dyadic-social-text/