--- title: "Joint Identification and Segmentation of Domain-Specific" description: "Conversational dialogue systems for new scenarios usually start with little NLU training data. 7 references, cited by 2 papers here." authors: ["Fabrizio Morbini", "Kenji Sagae"] year: 2011 venue: "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies" source: "https://aclanthology.org/P11-2017/" licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)" licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/morbini-2011-joint-identification-segmentation-domain-specific/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/morbini-2011-joint-identification-segmentation-domain-specific/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2011](https://md.research.elamur.ai/free/en/papers/2011/) / Joint Identification and Segmentation of Domain-Specific Dialogue Acts for Conversational Dialogue Systems # Joint Identification and Segmentation of Domain-Specific Dialogue Acts for Conversational Dialogue Systems **Authors:** Fabrizio Morbini and [Kenji Sagae](https://md.research.elamur.ai/free/en/authors/kenji-sagae/) · **Year:** 2011 · **Venue:** [Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2011](https://md.research.elamur.ai/free/en/venues/acl/2011/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/) **Links:** [Publisher's page](https://aclanthology.org/P11-2017/) · [PDF](https://aclanthology.org/P11-2017.pdf) · [OpenAlex](https://openalex.org/W2252233446) **Citations:** 37 citations (OpenAlex, 2026-10-14) · [2](https://md.research.elamur.ai/free/en/papers/morbini-2011-joint-identification-segmentation-domain-specific/cited-by/) in our library, of the 141,840 papers whose bibliographies we read ## Summary Conversational dialogue systems for new scenarios usually start with little NLU training data. Most collected examples express a single dialogue act, so longer user turns that combine several acts are poorly covered. Morbini and Sagae tackle this gap with a method that splits an utterance into consecutive, non-overlapping spans and labels each span with a domain-specific dialogue act. The method needs only a classifier trained on single-act examples. A maximum entropy classifier (the SASO NLU of Sagae et al., 2009) scores every possible word span. Dynamic programming then picks the segmentation with the best combined score, which costs O(n²) classifier calls. Tested on TACQ user data with 10-fold cross-validation, the method raises overall F-score above two baselines: a plain single-label classifier and a set of one-vs-rest binary classifiers. The gain is significant, and it is largest on multi-act utterances. Unlike the binary baseline, the method also reports which words carry each act. This weakens the chicken-and-egg loop: users who are understood poorly when they speak at length tend to shorten their turns, so multi-act data never accumulates. ## Key points - The TACQ dataset (Artstein et al., 2009) has 1,579 utterances. - 1,204 carry one act and 375 carry several, averaging 2.3 acts each. - Annotation uses 77 domain-specific labels such as whq(obj: strangeMan, attr: name), treated as atomic. - Training folds contain only single-act material: gold multi-act utterances are cut into their annotated segments. - Overall F-score is 75 for the method, 73 for the binary-classifier baseline and 71 for the single-label baseline (p \< 0.05). - On multi-act utterances the scores are 75, 67 and 55. - On single-act utterances the method drops slightly (75 vs 78 and 80). - To combine segment scores, the authors used a per-segment geometric mean rather than the raw product; the variants performed similarly. - Speed and boundary accuracy: - The method is about 25× slower than the single-label baseline and roughly 15% faster than the binary baseline. - It averages about 102 ms on multi-act utterances. - On multi-act data, predicted span edges miss by about ±2 words on average. ## Interesting ideas - A classifier trained only on whole short units can segment longer inputs: score every span, then let dynamic programming choose the best chain of spans, so no segmentation-annotated data is needed. - Normalizing the combined score per segment (a geometric mean instead of a raw product) removes the built-in preference for fewer, longer segments that products of probabilities create. - Training data for the simple case can be manufactured by cutting richer annotated examples into their parts, while the harder combined case is left to search at inference time. - The evaluation credits an output only if both its span and its label align with gold, and it reports a separate histogram of boundary offsets in words. - The comparison includes a one-vs-rest baseline that finds the right set of labels but cannot say where each one sits. This shows the value of localization beyond label-set accuracy. ## 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 split compound turns ("thanks, and where does he live?") into several acts using a recognizer trained only on single-act clauses. - **I2** (answer): when breaking a multi-part school question or word problem into sub-questions, a per-part normalized score would avoid unfairly favouring a single merged reading. - **I3** (learn): silver data of complex sentences could be split into single-frame clause examples for training, with composition handled by a decoder rather than learned directly. - **I4** (learn): frame and coreference evaluations could likewise score label correctness and span placement separately, exposing near-miss boundary errors that a single accuracy figure hides. - **I5** (frames): every detected speech act or frame should be tied to its word span, which keeps the pipeline explainable instead of yielding an unanchored set of labels. *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 - [University of Southern California](https://md.research.elamur.ai/free/en/universities/university-of-southern-california/) — University, Los Angeles, United States · 104 works · 33,876 citations 📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,085 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,888 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works 🧭 Topics (2): [🗨️ Dialogue systems](https://md.research.elamur.ai/free/en/topics/dialogue-systems/) · 401 works · [💬 Speech act theory](https://md.research.elamur.ai/free/en/topics/speech-act-theory/) · 160 works ## Cite Cite as: Fabrizio Morbini and Kenji Sagae. 2011. Joint Identification and Segmentation of Domain-Specific Dialogue Acts for Conversational Dialogue Systems. Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{morbini2011joint, title = {{Joint Identification and Segmentation of Domain-Specific Dialogue Acts for Conversational Dialogue Systems}}, author = {Fabrizio Morbini and Kenji Sagae}, booktitle = {Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies}, year = {2011}, url = {https://aclanthology.org/P11-2017/} } ``` - [← previous Semi-Supervised Frame-Semantic Parsing for Unknown Predicates](https://md.research.elamur.ai/free/en/papers/das-2011-semi-supervised-frame-semantic-parsing/) - [next → Bootstrapping coreference resolution using word associations](https://md.research.elamur.ai/free/en/papers/kobdani-2011-bootstrapping-coreference-resolution-word-associations/) - [↑ back to papers of 2011](https://md.research.elamur.ai/free/en/papers/2011/) ## References 7 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Ron Artstein, Sudeep Gandhe, Michael Rushforth, and David R. Traum. 2009. Viability of a simple dialogue act scheme for a tactical questioning dialogue system. In DiaHolmia 2009: Proceedings of the 13th Workshop on the Semantics and Pragmatics of Dialogue, page 43–50, Stockholm, Sweden, June. [find](https://scholar.google.com/scholar?q=Ron+Artstein%2C+Sudeep+Gandhe%2C+Michael+Rushforth%2C+and+David+R%2E+Traum%2E+2009%2E+Viability+of+a+simple+dialogue+act+scheme+for+a+tactical+questioning+dialogue+system%2E+In+DiaHolmia+2009%3A+Proceedings+of+the+13) 2. Oliver Bender, Klaus Macherey, Franz Josef Och, and Hermann Ney. 2003. Comparison of alignment templates and maximum entropy models for natural language understanding. In Proceedings of the tenth conference on European chapter of the Association for Computational Linguistics - Volume 1, EACL ’03, pages 11–18, Stroudsburg, PA, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Oliver+Bender%2C+Klaus+Macherey%2C+Franz+Josef+Och%2C+and+Hermann+Ney%2E+2003%2E+Comparison+of+alignment+templates+and+maximum+entropy+models+for+natural+language+understanding%2E+In+Proceedings+of+the+tenth+conf) 3. Sudeep Gandhe, David DeVault, Antonio Roque, Bilyana Martinovski, Ron Artstein, Anton Leuski, Jillian Gerten, and David R. Traum. 2008. From domain specification to virtual humans: An integrated approach to authoring tactical questioning characters. In Proceedings of Interspeech, Brisbane, Australia, September. [find](https://scholar.google.com/scholar?q=Sudeep+Gandhe%2C+David+DeVault%2C+Antonio+Roque%2C+Bilyana+Martinovski%2C+Ron+Artstein%2C+Anton+Leuski%2C+Jillian+Gerten%2C+and+David+R%2E+Traum%2E+2008%2E+From+domain+specification+to+virtual+humans%3A+An+integrated+appro) 4. Kenji Sagae, Gwen Christian, David DeVault, and David R. Traum. 2009. Towards natural language understanding of partial speech recognition results in dialogue systems. In Short Paper Proceedings of the North American Chapter of the Association for Computational Linguistics - Human Language Technologies (NAACL HLT) 2009 conference. [find](https://scholar.google.com/scholar?q=Kenji+Sagae%2C+Gwen+Christian%2C+David+DeVault%2C+and+David+R%2E+Traum%2E+2009%2E+Towards+natural+language+understanding+of+partial+speech+recognition+results+in+dialogue+systems%2E+In+Short+Paper+Proceedings+of+th) 5. Andreas Stolcke and Elizabeth Shriberg. 1996. Automatic linguistic segmentation of conversational speech. In Proc. ICSLP, pages 1005–1008. [find](https://scholar.google.com/scholar?q=Andreas+Stolcke+and+Elizabeth+Shriberg%2E+1996%2E+Automatic+linguistic+segmentation+of+conversational+speech%2E+In+Proc%2E+ICSLP%2C+pages+1005%E2%80%931008%2E) 6. David R. Traum, Anton Leuski, Antonio Roque, Sudeep Gandhe, David DeVault, Jillian Gerten, Susan Robinson, and Bilyana Martinovski. 2008a. Natural language dialogue architectures for tactical questioning characters. In Army Science Conference, Florida, 12/2008. [find](https://scholar.google.com/scholar?q=David+R%2E+Traum%2C+Anton+Leuski%2C+Antonio+Roque%2C+Sudeep+Gandhe%2C+David+DeVault%2C+Jillian+Gerten%2C+Susan+Robinson%2C+and+Bilyana+Martinovski%2E+2008a%2E+Natural+language+dialogue+architectures+for+tactical+question) 7. David R. Traum, Stacy Marsella, Jonathan Gratch, Jina Lee, and Arno Hartholt. 2008b. Multi-party, multi-issue, multi-strategy negotiation for multi-modal virtual agents. In IVA, pages 117–130. [find](https://scholar.google.com/scholar?q=David+R%2E+Traum%2C+Stacy+Marsella%2C+Jonathan+Gratch%2C+Jina+Lee%2C+and+Arno+Hartholt%2E+2008b%2E+Multi%2Dparty%2C+multi%2Dissue%2C+multi%2Dstrategy+negotiation+for+multi%2Dmodal+virtual+agents%2E+In+IVA%2C+pages+117%E2%80%93130%2E) *Updated 24 October 2026*