Elamur Research⌕
☰
home / papers / 2010 / Using Syntactic and Semantic based Relations for Dialogue Act Recognition

Using Syntactic and Semantic based Relations for Dialogue Act Recognition

Paper · 2010 · Tina Klüwer, Hans Uszkoreit and Feiyu Xu

Summary

The paper asks whether dialogue act recognition improves when a classifier sees structured relations from the utterance, not only its words. The setting is KomParse, a text-chat sales agent in the virtual world Twinity. The authors used a Wizard-of-Oz corpus in which 18 users furnished a virtual living room by chatting with a human-controlled furniture seller. They tagged it with seven domain-specific dialogue acts plus pragmatic features: sentence mood, topic, and the preceding act and topic. A rule-based parser then produced PropBank-style predicate argument structures from dependency trees, which were corrected by hand. A further manual layer mapped each predicate to a VerbNet class. A Bayesian network classifier gained clearly when these relations joined the context features, rising from 61.9% to 67.4% accuracy. A ROCCHIO-style classifier confirmed the same trend at a higher level, reaching 74.4%. The authors also found that a few predicates cover most utterances. This led them to add an automatically generated rule component for robustness. The work shows that compact predicate-argument and frame information carries intent signal beyond surface words.

Key points

  • Corpus: 18 Wizard-of-Oz dialogues containing 3,171 turns, 4,313 utterances and about 23,015 words.
  • Dialogue act set: seven acts in total. Two ask for something (REQUEST INFO and REQUEST), two respond to offers (REJECT and ACCEPT), and the remaining three are PROPOSE, ACKNOWLEDGE and PROVIDE INFO. REQUEST is by far the most frequent (449), and ACKNOWLEDGE is the rarest (9).
  • Bayesian network results:
    • Words only: 48.1%.
    • Best baseline: 61.9%.
    • With relations: 67.4%.
    • Partial sets alone: context only 56.6%, VerbNet relations only 53.5%, syntactic relations only 50.8%.
  • ROCCHIO results: utterances alone 70.1%, adding context 73.2%, adding syntactic relations 74.4%.
  • Predicate coverage: there are 80 distinct syntactic predicates and 66 VerbNet predicates. Just 21 predicate classes cover 1,101 of 1,239 instances. The dominant classes include want-32.1, own-100, reflexive_appearance-48.1.2, put-9.1 and see-30.1.

Interesting ideas

  • I1. Stable argument labels: instead of PropBank's verb-specific numbering, arguments get labels by syntactic position, so the subject is always arg1. This gives downstream components a consistent label meaning across verbs.
  • I2. Declarative extraction rules: the predicate-argument extractor is a set of declarative rules, each describing a dependency subtree by its grammatical functions. Relation extraction stays transparent and editable.
  • I3. Parallel representations as an ablation: the same utterances get a syntactic relation layer and a more abstract VerbNet layer, and each is tested separately. The more abstract layer alone mapped somewhat better to intents.
  • I4. Mining coverage to build rules: the analysis showed that a small set of domain predicates dominates the data. The authors exploited this by generating classification rules automatically from the annotated features, alongside the statistical classifier.
  • I5. Indirect requests: polite questions such as "can you show me" are treated as requests, using mood, modal and predicate together rather than surface punctuation.

For Mova

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

  • I1 (frames): prag could keep a verb-independent syntactic slot layer under the VerbNet roles, so speech-act rules can match "subject + modal + action" patterns without knowing each verb's role inventory.
  • I2 (tree): relation patterns over UD trees could be stored as explicit subtree templates, so every extracted relation can point back to the rule and nodes that produced it.
  • I3 (learn): evaluation could run the same downstream task once on raw syntactic frames and once on VerbNet-abstracted frames, to measure what each level of abstraction actually contributes.
  • I4 (learn): frequency analysis of frames in silver data could pick the high-coverage patterns worth turning into self-induced, inspectable rules, leaving the long tail to learned models.
  • I5 (frames): the speech-act detector could combine interrogative mood, a modal argM and an action predicate whose agent is the addressee to label indirect requests explainably.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Arts 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 Tina Klüwer, Hans Uszkoreit and Feiyu Xu. 2010. Using Syntactic and Semantic based Relations for Dialogue Act Recognition. Coling 2010: Posters.

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

@inproceedings{kluwer2010using,
  title = {{Using Syntactic and Semantic based Relations for Dialogue Act Recognition}},
  author = {Tina Klüwer, Hans Uszkoreit and Feiyu Xu},
  booktitle = {Coling 2010: Posters},
  year = {2010},
  url = {https://aclanthology.org/C10-2065/}
}

References

26 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Allen, James F., Bradford W. Miller, Eric K. Ringger, and Teresa Sikorski. 1996. A robust system for natural spoken dialogue. In Proceedings of ACL 1996. find
  2. Allen, James, Mehdi Manshadi, Myroslava Dzikovska, and Mary Swift. 2007. Deep linguistic processing for spoken dialogue systems. In DeepLP ’07: Proceedings of the Workshop on Deep Linguistic Processing, Morristown, NJ, USA. find
  3. Andernach, Toine. 1996. A machine learning approach to the classification of dialogue utterances. CoRR, cmp-lg/9607022. ↗
  4. Baker, Collin F., Charles J. Fillmore, and John B. Lowe. 1998. The berkeley framenet project. In Proceedings of COLING 1998.
  5. Bertomeu, Nuria and Anton Benz. 2009. Annotation of joint projects and information states in human-npc dialogues. In Proceedings of CILC-09, Murcia, Spain. find
  6. Clark, H.H. 1996. Using Language. Cambridge University Press. find
  7. de Marneffe, Marie C. and Christopher D. Manning. 2008. The Stanford typed dependencies representation. In Coling 2008: Proceedings of the workshop on Cross-Framework and Cross-Domain Parser Evaluation, Manchester, UK.
  8. Jurafsky, Daniel, Elizabeth Shriberg, Barbara Fox, and Traci Curl. 1998. Lexical, prosodic, and syntactic cues for dialog acts.
  9. Keizer, Simon and Rieks op den Akker. 2006. Dialogue act recognition under uncertainty using bayesian networks. Nat. Lang. Eng., 13(4). find
  10. Keizer, Simon, Rieks op den Akker, and Anton Nijholt. 2002. Dialogue act recognition with bayesian networks for dutch dialogues. In Proceedings of the 3rd SIGdial workshop on Discourse and dialogue, Morristown, NJ, USA.
  11. Klüwer, Tina, Peter Adolphs, Feiyu Xu, Hans Uszkoreit, and Xiwen Cheng. 2010. Talking npcs in a virtual game world. In Proceedings of the System Demonstrations Section at ACL 2010. find
  12. Lapata, Mirella and Alex Lascarides. 2004. Inferring sentence-internal temporal relations. In Proceedings of the North American Chapter of the Association for Computational Linguistics, pages 153–160.
  13. Neumann, Günter and Sven Schmeier. 2002. Shallow natural language technology and text mining. Künstliche Intelligenz. The German Artificial Intelligence Journal. find
  14. Palmer, Martha, Daniel Gildea, and Paul Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. Comput. Linguist., 31(1). ↗
  15. Schmid, Helmut. 1994. In Proceedings of the International Conference on New Methods in Language Processing. find
  16. Schuler, Karin Kipper. 2005. Verbnet: a broad-coverage, comprehensive verb lexicon. Ph.D. thesis, Philadelphia, PA, USA. ↗
  17. Searle, John R. 1969. Speech acts : an essay in the philosophy of language / John R. Searle. Cambridge University Press, London. ↗
  18. Sporleder, Caroline and Alex Lascarides. 2008. Using automatically labelled examples to classify rhetorical relations: A critical assessment. Natural Language Engineering, 14(3). find
  19. Stolcke, Andreas, Klaus Ries, Noah Coccaro, Elizabeth Shriberg, Rebecca Bates, Daniel Jurafsky, Paul Taylor, Rachel Martin, Carol Van, and Ess dykema Marie Meteer. 2000. Dialogue act modeling for automatic tagging and recognition of conversational speech. Computational Linguistics, 26.
  20. Subba, Rajen and Barbara Di Eugenio. 2009. An effective discourse parser that uses rich linguistic information. In NAACL ’09, Morristown, NJ, USA.
  21. Surendran, Dinoj and Gina-Anne Levow. 2006. Dialog act tagging with support vector machines and hidden markov models. In Interspeech. find
  22. Verbree, A.T., R.J. Rienks, and D.K.J. Heylen. Dialogue-act tagging using smart feature selection: results on multiple corpora. In Raorke, B., editor, First International IEEE Workshop on Spoken Language Technology SLT 2006. find
  23. Webb, Nick and Ting Liu. 2008. Investigating the portability of corpus-derived cue phrases for dialogue act classification. In Proceedings of COLING 2008, Manchester, UK. find
  24. Xu, Feiyu, Hans Uszkoreit, and Hong Li. 2007. A seed-driven bottom-up machine learning framework for extracting relations of various complexity. In Proceedings of ACl (07), Prague, Czech Republic. find
  25. Zheng, Fei and Geoffrey I. Webb. 2006. Efficient lazy elimination for averaged one-dependence estimators. In ICML, pages 1113–1120. find
  26. Zimmermann, Matthias, Yang Liu, Elizabeth Shriberg, and Andreas Stolcke. 2005. Toward joint segmentation and classification of dialog acts in multiparty meetings. In Proc. Multimodal Interaction and Related Machine Learning Algorithms Workshop (MLMI05, page 187. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/kluwer-2010-syntactic-semantic-based-relations-dialogue/