Elamur Research⌕
☰
home / papers / 2015 / Arabic Inquiry-Answer Dialogue Acts Annotation Schema

Arabic Inquiry-Answer Dialogue Acts Annotation Schema

Paper · 2015 · AbdelRahim A. Elmadany, Sherif M. Abdou and Mervat Gheith

Summary

The paper addresses the lack of dialogue act (DA) resources for Arabic customer-service conversations. Earlier Arabic schemes targeted general discussion such as TV talk shows, labelled individual words rather than utterances, or used tag sets too narrow to cover suggestions, warnings, corrections and social obligations. The authors build on English schemes (MapTask, VERBMOBIL, DAMSL, SWITCHBOARD, DIT++, ISO 24617-2) and earlier Arabic work to design a schema for inquiry-answer dialogue. It has three dimensions: request acts, response acts, and "other" acts that handle opening, closing and self-introduction. They apply it to a new, manually segmented and labelled corpus of Arabic dialect dialogues. The corpus combines transcribed phone calls with Egyptian banks and Egypt Air and written chats with Saudi mobile operators. It is presented as the first such annotated Arabic corpus, intended as a basis for Arabic language understanding in service-oriented dialogue systems. The paper reports no inter-annotator agreement and no classification experiments.

Key points

  • Schema structure: the text describes 6 request acts, 14 response acts and 3 other acts.
    • The tables actually list 7 request acts, including Taking-Request and Turn-Assign.
    • They also list 15 response acts, including Service-Answer, Agree/Disagree, MissUnderstandingSign, Correct, Pausing, Suggest, Promise, Warning and Offer.
  • Corpus size: the corpus is said to contain 83 dialogues, but the stated parts add up to 82.
    • 52 spoken calls total about two hours of talk after ads were removed; they were transcribed with Transcriber.
    • 30 chats come from KSA Zain, KSA Mobily and KSA STC.
  • Corpus statistics: there are 3001 turns, averaging 6.7 words per turn, and 4727 utterances, averaging 4.3 words per utterance.
  • Annotation tool: a custom tool was built so annotators could split turns into utterances and assign a DA label to each.
  • Motivation from prior work: the related-work section notes that SWITCHBOARD-DAMSL's 220 tags were impractical, with 130 of them occurring fewer than 10 times. That set was therefore reduced to 42 mutually exclusive types.

Interesting ideas

  • I1. Agree and Disagree are defined as polarity labels over whatever act came before, so accepting a confirmation, a yes-answer, thanks or an apology all share one tag; this keeps the tag set small while the meaning stays recoverable from the antecedent act.
  • I2. A whole turn can carry a composite label (Opening) while its segments get finer acts (Greeting, Self-Introduce, Self-Introduce), giving a two-level turn/utterance annotation.
  • I3. Questions are typed by the kind of answer they expect (yes/no, choice among options, confirmation, service information versus off-task details such as a phone number), and each response act names which question types it can satisfy.
  • I4. The SWITCHBOARD lesson is that a fine tag set with many rare classes hurts both annotation consistency and statistical modelling, so tags get merged until each class has enough examples.
  • I5. Interaction-management acts such as Pausing ("just a moment"), Turn-Assign and MissUnderstandingSign are kept separate from content acts.

For Mova

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

  • I1 (frames): speech-act frames could store a response as a polarity plus a link to the act it answers, instead of multiplying accept-X/reject-X labels.
  • I2 (frames): prag could assign a discourse-level act to a sentence group while keeping clause-level speech acts underneath, so both granularities are explainable.
  • I3 (answer): QA readers could classify a question's expected answer shape first and then check that the produced answer matches that shape, a cheap consistency test.
  • I4 (learn): when inducing labels from silver data, track per-class frequency and merge or back off rare labels before training, reporting the mapping so it stays explainable.
  • I5 (frames): tagging such utterances as non-propositional would stop them from injecting spurious events or states into the world of the text.

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 3Research subjects and languages the work is about.Semitic languages · 2,908 worksDialogue systems · 624 worksSpeech act theory · 229 works

Cite

Cite as AbdelRahim A. Elmadany, Sherif M. Abdou and Mervat Gheith. 2015. Arabic Inquiry-Answer Dialogue Acts Annotation Schema. arXiv:1505.04197.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1505.04197 · arXiv 1505.04197

@misc{elmadany2015arabic,
  title = {{Arabic Inquiry-Answer Dialogue Acts Annotation Schema}},
  author = {AbdelRahim A. Elmadany, Sherif M. Abdou and Mervat Gheith},
  year = {2015},
  doi = {10.48550/arxiv.1505.04197},
  eprint = {1505.04197},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1505.04197}
}

References

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

  1. [3]. A. Stolcke, K. Ries, N. Coccaro, E. Shriberg, R. Bates, D. Jurafsky, P. Taylor, R. Martin, C. V. Ess-Dykema, and M. Meteer, "Dialogue Act Modeling for Automatic Tagging and Recognition of Conversational Speech," Computational Linguistics, vol. 26, pp. 339-373, 2000. ↗
  2. [4]. A. Anderson, M. Bader, E. Bard, E. Boyle, and G. Doherty, "The HCRC Map Task Corpus," Language and Speech, vol. 34, pp. 351-366, 1991. ↗
  3. [5]. W. Wahlster, Verbmobil: Foundations of Speech-To-Speech Translation: Springer, 2000. find
  4. [6]. S. Jekat, A. Klein, E. Maier, Ilona Maleck, M. Mast, and J. Quantz, "Dialogue Acts in VERBMOBIL. VM report 65," Universities of Berlin, Saarbrcken and Stuttgart.1995. find
  5. [7]. J. Alexandersson, B. Buschbeck-Wolf, T. Fujinami, M. Kipp, S. Koch, E. Maier, N. Reithinger, B. S. mitz, and M. Siegel., "Dialogue Acts in VERBMOBIL- 2 (second edition). VM report 226 , DFKI GmbH, ," Universities of Berlin, Saarbrcken and Stuttgart1998. [8]. J. Allen and M. Core, "Draft of DAMSL: Dialog Act Markup in Several Layers," 1997. find
  6. [9]. M. Core and J. Allen, "Coding Dialogs with the DAMSL annotation scheme," presented at the AAAI Fall Symposium on Communicative Action in Humans and Machines, MIT, Cambridge, MA, 1997. find
  7. [10]. M. Core, M. Ishizaki, J. Moore, C. Nakatani, N. Reithinger, D. Traum, and S. Tutiya, "Report of The Third Workshop of the Discours e Resource Initiative," Chiba University, Chiba, Japan1998. [11]. D. Jurafsky, R. Bates, N. Coccaro, R. Martin, M. Meteer, K. Ries, E. Shriberg, A. Stolcke, P. Taylor, and C. E. Dykema., "Automatic Detection of Discourse Structure for Speech Recognition and Understanding," presented at the IEEE Workshop on Speech Recognition and Understanding, Santa Barbara, 1997. find
  8. [12]. D. Jurafsky, R. Bates, N. Coccaro, R. Martin, M. Meteer, K. Ries, E. Shriberg, A. Stolcke, P. Taylor, and C. V. Ess-Dykema, "Switch-board Discourse Language Modeling Project Final Report," in ResearchNote 30, Center for Language and Speech Processing, Johns Hopkins University, Baltimore, 1998. find
  9. [13]. N. Webb, "Cue-Based Dialogue Act Classification," Ph.D. dissertation, Department of Computer Science, University of Sheffield, England, 2010. find
  10. [14]. E. Shriberg, R. Dhillon, S. Bhagat, J. Ang, and H. Carvey, " The ICSI Meeting Recorder Dialog Act (MRDA) Corpus," in Special Interest Group on Discourse and Dialogue (SIGdial), Boston, USA, 2004. ↗
  11. [15]. H. Bunt, "Context and Dialogue Control. , 1," THINK, vol. 3, pp. 19-31, 1994. find
  12. [16]. H. Bunt and Y. Girard, "Designing an Open, Multidimensional Dialogue Act Taxonomy," in the 9th Workshop on the Semantics and Pragmatics in Dialogue, 2005. find
  13. [17]. H. Bunt, J. Alexandersson, J. Carletta, J.-W. Choe, A. C. Fang, K. Hasida, K. Lee, V. Petukhova, A. Popescu-Belis, L. Romary, C. Soria, and D. Traum, "Towards an ISO standard for dialogue act annotation," in the International Conference on Language Resources and Systems (LREC 2010), Malta, Italy, 2010, pp. 2548–2558.
  14. [18]. H. Bunt, J. Alexandersson, J. Carletta, J.-W. Choe, A. C. Fang, K. Hasida, K. Lee, V. Petukhova, A. Popescu-Belis, L. Romary, C. Soria, and D. Traum, "ISO 24617-2: A semantically-based standard for dialogue annotation," in Proceedings of the International Conference on Language Resources and Evaluation (LREC), Istanbul, Turkey, 2012.
  15. [19]. L. Shala, V. Rus, and A. Graesser, "Automatic Speech Act Classification In Arabic," in Subjetividad y Procesos Cognitivos Conference 2010, pp. 284-292. find
  16. [20]. S. B. Dbabis, F. Mallek, H. Ghorbel, and L. Belguith, "Dialogue Acts Annotation Scheme within Arabic discussion," in SemDial 2012 The 16th workshop on semantics and pragmatics of dialogue Sorbonne, Paris, France, 2012. find
  17. [21]. M. Graja, M. Jaoua, and L. H. Belguith, "Discriminative Framework for Spoken Tunisian Dialect Understanding," presented at th e SLSP, 2013. find
  18. [22]. C. Lhioui, A. Zouaghi, and M. Zrigui, "A Combined Method Based on Stochastic and Linguistic Paradigm for the Understanding of Arabic Spontaneous Utterances," in CICLing 2013, Computational Linguistics and Intelligent Text Processing Lecture Notes in Computer Science, Samos, Greece, 2013, pp. 549-558. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/elmadany-2015-arabic-inquiry-answer-dialogue-acts/