Tagging of Speech Acts and Dialogue Games in Spanish Call Home

Summary

This paper reports early work from Carnegie Mellon's Clarity project. The project's goal is to recognize discourse structure in casual conversation that has no shared task, using shallow corpus-based techniques. A longer-term aim is to detect functional activities such as planning or gossiping. The authors annotated the Call Home Spanish (CHS) corpus of family phone calls at two levels. Speech acts were tagged with an adapted SWBD-DAMSL scheme. Dialogue games were tagged with an inventory redesigned for open-ended talk rather than Map Task-style collaboration.

They then applied a multi-level classifier based on n-gram and recursive HMM models over segment lattices. Instead of handling discourse levels one after another, it uses speech acts as an intermediate layer between words and games. In preliminary tests, word n-grams could not predict game labels. Speech act sequences beat the baseline, but only when the speech acts were hand-annotated rather than automatically recognized. The work contributes a tagged Spanish conversational resource. It also shows that higher-level discourse classification needs linguistic cues beyond current speech act tags.

Key points

Speech act tagset changes: After two coders reached 79% agreement, the opinion tag sv was merged into plain statement s because coders could not separate the two reliably. Semantic features were added to statements, including certainty, hypotheticality, val+/- judgements, obligation, tense, and joke/sarcasm. An attention-directive tag (atd) was added for cues like oye and mira. One annotator then retagged the whole corpus, with intra-coder agreement above 90%.
Statement subtypes: Adding features cut unmarked s from roughly 40% to 26% of tags. However, the classifier could not reliably tell the marked subtypes apart from plain s.
Game scheme: It has eight game types (e.g., seeking/giving information, giving directive, action commit, expressive, communication filler) and eight modifiers (e.g., elab, spec, n, inc, aband). Turns are labelled Initiative, Response, or Feedback, and games may be nested or interleaved.
Architecture: An A* search segments a word lattice, and the segments are relabelled as speech acts and then games. A Viterbi search selects the best joint path. The same framework is also used for dialogue act detection in a travel-domain speech translation system.

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: Discourse analysis · Speech act theory

Cited by

6 papers of our library cite it (of the 140673 whose bibliographies we read); 1 of them have a page here, 5 not yet.

Dialogue Act Tagging for Instant Messaging Chat Sessions

Cite

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

@inproceedings{levin1999tagging,
  title = {{Tagging of Speech Acts and Dialogue Games in Spanish Call Home}},
  author = {Levin, Lori and Ries, Klaus and Thyme-Gobbel, Ann and Lavie, Alon},
  booktitle = {Towards Standards and Tools for Discourse Tagging},
  year = {1999},
  doi = {10.5445/ir/1000166527},
  url = {https://doi.org/10.5445/ir/1000166527}
}

References

11 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

Douglas Biber, Susan Conrad, and Randi Reppen. Corpus Lingusitics: Investigating Language Structure and Use. Cambridge University Press, 1998. find
Jean Carletta, Amy Isard, Stephen Isard, Jacqueline C. Kowtko, Gwyneth Doherty-Sneddon, and Anne H. Anderson. The reliability of a dialogue structure coding scheme. Computational Linguistics, 23(1): 13-31, March 1997. find
Suzanne Eggins and Diana Slade. Analysing • Casual Conversation. Cassell, 1997. find
Michael Finke, Maria Lapata, Alon Lavie, Lori Levin, Laura Mayfield Tomokiyo, Thomas Polzin, Klaus Ries, Alex Waibel, and Klaus Zechner. Clarity: Automatic discourse and dialogue analysis for a speech and natural language processing system. In AAAI Spring Symposium on Applying Machine Learning to Di… find
Dan Jurafsky, Liz Shriberg, and D. Biasca. Switchboard-damsl labeling project coder's manual. Technical report, Institute of Cognitive Science, University of Colorado, Boulder, USA, 1997, find
Lori Levin, Ann Thym~-Gobbel, Klaus Ries, Alon Lavie, and Monika Woszczyna. A discourse coding scheme for conversational spanish. In ICSLP, 1998. find
Klaus Ries. Hmm and neural network based speech act detection. In ICASSP, 1999. find
Andreas Stolcke, Elizabeth Shriberg, Rebecca Bates, Noah Coccaro, Daniel Jurafsky, Rachel Martin, Marie Meteer, Klaus Ries, Paul Taylor, and Carol Van Ess-Dykema. Dialog act modeling for conversational speech. In AAAI Spring Symposium on Applying Machine Learning to Discourse Processing, March 1998. find
Ann Thym~-Gobbel and Lori Levin. Dialogue Activity Tagging Manual for Spanish Conversational Speech. http ://wwW • cnbc. cmu. edu/'gobbel/clar ity/ manual intro, html, 1998. find
V. Warnke, S. Harbeck, H. Niemann, and E. NSth. Discriminative estimation of interpolation parameters for language model clas-sifters. In ICASSP, 1999. find
M. Woszczyna, M. Broadhead, D. Gates, M. GavaldA, A. Lavie, L. Levin, and A. Waibel. A modular approach to spoken language translation for large domains. In AMTA-98, 1998. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/levin-1999-tagging-speech-acts-dialogue-games/

📄 Tagging of Speech Acts and Dialogue Games in Spanish Call Home