Using Speech Acts in Logic-Based Rhetorical Structuring for Natural Language Generation in Human-Computer Dialogue

What it is about

A task-oriented spoken dialogue system builds the discourse structure of its turns with a first-order logic version of Segmented Discourse Representation Theory (SDRT). Each new utterance is checked against each earlier one for all 17 rhetorical relations, which costs N × R proofs, where N is the number of utterances and R = 17. The authors attach a speech-act type to each utterance (for example an inform or a request for information) and use corpus-derived tables that allow only about M = 3 relations per pair of speech acts. Checking only those relations cuts the cost by about R/M, roughly six times on their averages. The speech-act constraint also removes relations that the logic accepts but that do not fit the dialogue, such as a background relation between an answer and an earlier question. The paper is a short system description and reports no large evaluation.

Ideas

• I1. A table from pairs of speech-act types to allowed discourse relations, stored in a BTreeMap in prag, and consulted before any relation check. Count the proofs with and without the table on the dialogue examples.
• I2. Keep the cost measure: proofs per new utterance, N × M against N × R, reported for each dialogue in a test set.
• I3. Negative control: replace the table with a random subset of relations of the same size. The accepted relations should then disagree with the intended ones more often.
• I4. Use the same table for the discourse relations of school texts, with speech-act tags from the frame reader in prag/src/frames.rs.

For Mova

Chapters: 31. Speech acts, 36. Discourse relations and coherence, 38. Dialogue, discourse markers and reported speech. Code: speech-act tagging and frames in prag, relation checks in the same crate. Measure: proofs per utterance and the agreement of accepted relations with a gold label, against the random-subset control; two runs must match.

Topics: Discourse analysis · Speech act theory

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/2007.sigdial-1.44

@inproceedings{popescu2007using,
  title = {{Using Speech Acts in Logic-Based Rhetorical Structuring for Natural Language Generation in Human-Computer Dialogue}},
  author = {Popescu, Vladimir and Caelen, Jean and Burileanu, Corneliu},
  booktitle = {Proceedings of the 8th SIGdial Workshop on Discourse and Dialogue},
  year = {2007},
  doi = {10.18653/v1/2007.sigdial-1.44},
  url = {https://doi.org/10.18653/v1/2007.sigdial-1.44}
}

References

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

1. N. Asher and A. Lascarides. 2003. Logics of Conversation. Cambridge University Press. find
2. J. Caelen and A. Xuereb. 2007. Interaction et Pragmatique - jeux de dialogue et de langage. Editions Hermès-Lavoisier, Paris. find
3. L. Danlos, B. Gaiffe and L. Roussarie. 2003. Document Structuring à la SDRT. International Workshop on Generation, ACL, Toulouse. find
4. M. F. McTear. 2002. Spoken Language Technology: Enabling the Conversational User Interface. ACM Computer Surveys, 34(1). find
5. V. Popescu, J. Caelen and C. Burileanu. 2007. Logic-Based Rhetorical Structuring Component in Natural Language Generation for Human-Computer Dialogue. find
6. Text, Speech and Dialog 2007, Pilsen, Czech Republic, LNCS, Springer. find
7. A. Stent. 2002. A Conversation Acts Model for Generating Spoken Dialogue Contributions. Computer Speech and Language, 16. find
8. D. Vanderveken. 1990-1991. Meaning and Speech Acts. Cambridge University Press. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/popescu-2007-speech-acts-logic-based-rhetorical/