---
title: "Modeling Dialogue Acts with Content Word Filtering and Speaker"
description: "The authors propose CSM, an HMM-based generative model with two main additions. 27 references, cited by 3 papers here."
authors: ["Yohan Jo", "Michael Yoder", "Hyeju Jang", "Carolyn Rosé"]
year: 2017
venue: "Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing"
source: "https://aclanthology.org/D17-1232/"
doi: "10.18653/v1/d17-1232"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/jo-2017-modeling-dialogue-acts-content-word/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/jo-2017-modeling-dialogue-acts-content-word/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2017](https://md.research.elamur.ai/free/en/papers/2017/page/5/) / Modeling Dialogue Acts with Content Word Filtering and Speaker Preferences

# Modeling Dialogue Acts with Content Word Filtering and Speaker Preferences

**Authors:** [Yohan Jo](https://md.research.elamur.ai/free/en/authors/yohan-jo/), [Michael Yoder](https://md.research.elamur.ai/free/en/authors/michael-yoder/), Hyeju Jang and [Carolyn Rosé](https://md.research.elamur.ai/free/en/authors/carolyn-penstein-rose/) · **Year:** 2017 · **Venue:** [Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing](https://md.research.elamur.ai/free/en/venues/emnlp/) ([EMNLP 2017](https://md.research.elamur.ai/free/en/venues/emnlp/2017/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/D17-1232/) · [PDF](https://aclanthology.org/D17-1232.pdf) · [DOI 10.18653/v1/d17-1232](https://doi.org/10.18653/v1/d17-1232) · [OpenAlex](https://openalex.org/W2756640946)

**Citations:** 8 citations (OpenAlex, 2026-10-14) · 3 in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper tackles unsupervised recognition of dialogue acts (DAs) in conversations. The authors propose CSM, an HMM-based generative model with two main additions. First, it filters content words. Each conversation receives a slowly changing background topic, drawn from topics shared across the corpus, which absorbs subject-matter vocabulary. Foreground topics then keep the function-like words that signal DAs. Second, it models speaker preferences: the DA of an utterance depends on both the parent utterance's state and the speaker's own tendencies.

CSM also treats an utterance-level DA as a mixture of sentence-level acts and follows reply trees without duplicating posts. It was evaluated on two very different corpora, the CNET forum and the NPS Chat corpus. Content filtering helped mainly on CNET, while speaker preferences helped mainly on NPS. Because the components compensate for one another, CSM stayed robust on both corpora and beat recent baselines. This matters because it shows which modeling assumptions suit which kinds of conversation, and that function words carry most of the DA signal.

## Key points
- CSM stands for content word filtering and speaker preferences model. It extends an HMM: each hidden state is an utterance-level DA, and each sentence is assigned one foreground topic representing a finer-grained act.
- Each word is labeled foreground or background through a switch. Background topics are learned without domain labels, unlike Lee et al. (2013), and no general background language model is used.
- The state of an utterance is drawn from a weighted mix of the parent's transition distribution and the speaker's preference distribution, with mixing weight ν.
  - The best setting was ν = 0.58 on NPS and ν = 1 on CNET, meaning CNET gets no speaker contribution.
  - Speakers are treated as conversation-specific.
- On NPS, removing speaker preferences lowered v-measure by 0.13.
- Supporting multi-level (reply-tree) structure improved accuracy on both corpora. For NPS, the structure was inferred from user mentions using simple heuristics.
- CSM was compared against six baselines, listed from oldest to newest: Ritter et al. (2010), Joty et al. (2011), Paul (2012), Lee et al. (2013), Ezen-Can and Boyer (2015), and Brychcín and Král (2017).

## Interesting ideas
- Separating content from function by how fast each changes over time, with topics drifting slowly and acts changing quickly, gives a label-free way to isolate words that signal intent from words about the subject.
- Each sentence gets exactly one fine-grained act, and an utterance is a mixture of those acts, which models multi-act turns without assigning acts word by word.
- An explicit, tunable weight balances "what usually follows the previous act" against "what this speaker tends to do." Tuning it per corpus revealed when speaker style actually matters.
- Transitions are defined on parent–child pairs of a reply tree rather than a flat sequence. This handles branching threads without duplicating utterances or merging conflicting labels later.
- Ablating each component on deliberately dissimilar corpora shows which assumption helps under which data property, such as long versus short turns or a stable topic.

## For Mova

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

- **I1** (frames): Speech-act detection in prag could score cue words such as discourse markers, modals and light verbs separately from the document's topical lemmas, so each act decision can cite only its function-word evidence.
- **I2** (frames): A turn- or paragraph-level speech act could be built from per-sentence acts (e.g., explain situation + ask + thank), keeping the composition explicit and traceable.
- **I3** (world): Each discourse participant entity could store a sparse profile of preferred acts, used as an inspectable prior when interpreting ambiguous utterances in dialogues, fables or tales.
- **I4** (world): Event and utterance ordering in the text world could be a tree of reply links rather than a single line, so each act's antecedent is unambiguous.
- **I5** (learn): Evaluation could include per-genre ablations of pragmatic components, reporting where each rule or prior helps or hurts instead of a single pooled score.

*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

- [Carnegie Mellon University](https://md.research.elamur.ai/free/en/universities/carnegie-mellon-university/) — University, Pittsburgh, United States · 539 works · 179,533 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [🗨️ Dialogue systems](https://md.research.elamur.ai/free/en/topics/dialogue-systems/) · 624 works · [💬 Speech act theory](https://md.research.elamur.ai/free/en/topics/speech-act-theory/) · 229 works

## Cite

Cite as: Yohan Jo, Michael Yoder, Hyeju Jang and Carolyn Rosé. 2017. Modeling Dialogue Acts with Content Word Filtering and Speaker Preferences. Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing.

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

```bibtex
@inproceedings{jo2017modeling,
  title = {{Modeling Dialogue Acts with Content Word Filtering and Speaker Preferences}},
  author = {Yohan Jo, Michael Yoder, Hyeju Jang and Carolyn Rosé},
  booktitle = {Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing},
  year = {2017},
  doi = {10.18653/v1/d17-1232},
  url = {https://doi.org/10.18653/v1/d17-1232}
}
```

- [← previous Automatic Negation and Speculation Detection in Veterinary Clinical Text](https://md.research.elamur.ai/free/en/papers/cheng-2017-automatic-negation-speculation-detection-veterinary/)
- [next → BuzzSaw at SemEval-2017 Task 7: Global vs. Local Context for Interpreting and Locating Homographic English Puns with Sense Embeddings](https://md.research.elamur.ai/free/en/papers/oele-2017-buzzsaw-semeval-2017-task-7/)
- [↑ back to papers of 2017](https://md.research.elamur.ai/free/en/papers/2017/page/5/)

## References

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

1. D Scott Appling, Erica J Briscoe, Heather Hayes, and Rudolph L Mappus. 2013. Towards automated personality identification using speech acts. In AAAI Workshop - Technical Report. [find](https://scholar.google.com/scholar?q=D+Scott+Appling%2C+Erica+J+Briscoe%2C+Heather+Hayes%2C+and+Rudolph+L+Mappus%2E+2013%2E+Towards+automated+personality+identification+using+speech+acts%2E+In+AAAI+Workshop+%2D+Technical+Report%2E)
2. John L Austin. 1975. How to Do Things with Words. Harvard University Press. [↗](https://doi.org/10.1093/acprof:oso/9780198245537.001.0001)
3. Lee Becker, Wayne H Ward, Sarel van Vuuren, and Martha Palmer. 2011. DISCUSS: a dialogue move taxonomy layered over semantic representations. Proceedings of the Ninth International Conference on Computational Semantics, pages 310–314. [find](https://scholar.google.com/scholar?q=Lee+Becker%2C+Wayne+H+Ward%2C+Sarel+van+Vuuren%2C+and+Martha+Palmer%2E+2011%2E+DISCUSS%3A+a+dialogue+move+taxonomy+layered+over+semantic+representations%2E+Proceedings+of+the+Ninth+International+Conference+on+Compu)
4. Tomáš Brychcı́n and Pavel Král. 2017. [Unsupervised dialogue act induction using gaussian mixtures](https://md.research.elamur.ai/free/en/papers/brychcin-2017-unsupervised-dialogue-act-induction-gaussian/). In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 485–490, Valencia, Spain. Association for Computational Linguistics.
5. Gijsbert Erkens and Jeroen Janssen. 2008. Automatic coding of dialogue acts in collaboration protocols. International Journal of Computer-Supported Collaborative Learning, 3(4):447–470. [↗](https://doi.org/10.1007/s11412-008-9052-6)
6. Aysu Ezen-Can and Kristy Elizabeth Boyer. 2015. Understanding Student Language: An Unsupervised Dialogue Act Classification Approach. JEDM -Journal of Educational Data Mining, 7(1):51–78. [↗](https://doi.org/10.5281/zenodo.3554707)
7. Eric N Forsythand and Craig H Martell. 2007. Lexical and Discourse Analysis of Online Chat Dialog. In International Conference on Semantic Computing (ICSC 2007), pages 19–26. [find](https://scholar.google.com/scholar?q=Eric+N+Forsythand+and+Craig+H+Martell%2E+2007%2E+Lexical+and+Discourse+Analysis+of+Online+Chat+Dialog%2E+In+International+Conference+on+Semantic+Computing+%28ICSC+2007%29%2C+pages+19%E2%80%9326%2E)
8. Marsal Gavaldà. 2004. Soup: A Parser for Real-World Spontaneous Speech. Springer Netherlands, Dordrecht. [↗](https://doi.org/10.1007/1-4020-2295-6_17)
9. Shafiq Joty, Giuseppe Carenini, and Chin-Yew Lin. 2011. Unsupervised Modeling of Dialog Acts in Asynchronous Conversations. In Proceedings of the Twenty-Second International Joint Conference on Artificial Intelligence - Volume Volume Three, pages 1807–1813. AAAI Press. [find](https://scholar.google.com/scholar?q=Shafiq+Joty%2C+Giuseppe+Carenini%2C+and+Chin%2DYew+Lin%2E+2011%2E+Unsupervised+Modeling+of+Dialog+Acts+in+Asynchronous+Conversations%2E+In+Proceedings+of+the+Twenty%2DSecond+International+Joint+Conference+on+Artifi)
10. Daniel Jurafsky, Rebecca Bates, N. Coccaro, R. Martin, M. Meteer, K. Ries, E. Shriberg, A. Stolcke, Paul Taylor, and C. Van Ess-Dykema. 1998. Automatic detection of discourse structure for speech recognition and understanding. 1997 IEEE Workshop on Automatic Speech Recognition and Understanding Proceedings, pages 88–95. [↗](https://doi.org/10.5445/ir/1000166547)
11. Su Nam Kim, Li Wang, and Timothy Baldwin. 2010. Tagging and Linking Web Forum Posts. In Proceedings of the Fourteenth Conference on Computational Natural Language Learning, pages 192–202, Uppsala, Sweden. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Su+Nam+Kim%2C+Li+Wang%2C+and+Timothy+Baldwin%2E+2010%2E+Tagging+and+Linking+Web+Forum+Posts%2E+In+Proceedings+of+the+Fourteenth+Conference+on+Computational+Natural+Language+Learning%2C+pages+192%E2%80%93202%2C+Uppsala%2C+Swe)
12. Donghyeon Lee, Minwoo Jeong, Kyungduk Kim, Seonghan Ryu, and Gary Geunbae Lee. 2013. Unsupervised Spoken Language Understanding for a Multi-Domain Dialog System. IEEE Transactions on Audio, Speech, and Language Processing, 21(11):2451–2464. [↗](https://doi.org/10.1109/tasl.2013.2280212)
13. Stephen C Levinson. 1983. Conversational structure. In Pragmatics, chapter 6, pages 284–333. Cambridge University Press. [find](https://scholar.google.com/scholar?q=Stephen+C+Levinson%2E+1983%2E+Conversational+structure%2E+In+Pragmatics%2C+chapter+6%2C+pages+284%E2%80%93333%2E+Cambridge+University+Press%2E)
14. Diane J Litman and Katherine Forbes-Riley. 2006. Correlations between dialogue acts and learning in spoken tutoring dialogues. Natural Language Engineering, 12(02):161–176. [↗](https://doi.org/10.1017/s1351324906004165)
15. Max Louwerse, Art Graesser, Andrew Olney, and the Tutoring Research Group. 2002. Good Computational Manners: Mixed-Initiative Dialog in Conversational Agents. Etiquette for Human-Computer Work: Papers from the AAAI Fall Symposium, pages 71–76. [find](https://scholar.google.com/scholar?q=Max+Louwerse%2C+Art+Graesser%2C+Andrew+Olney%2C+and+the+Tutoring+Research+Group%2E+2002%2E+Good+Computational+Manners%3A+Mixed%2DInitiative+Dialog+in+Conversational+Agents%2E+Etiquette+for+Human%2DComputer+Work%3A+Papers)
16. J R Martin and David Rose. 2003. Negotiation: interacting in dialogue. New Century Series. Bloomsbury Academic. [find](https://scholar.google.com/scholar?q=J+R+Martin+and+David+Rose%2E+2003%2E+Negotiation%3A+interacting+in+dialogue%2E+New+Century+Series%2E+Bloomsbury+Academic%2E)
17. Amita Misra and Marilyn Walker. 2013. Topic Independent Identification of Agreement and Disagreement in Social Media Dialogue. Proceedings of the SIGDIAL 2013 Conference, (August):41–50. [find](https://scholar.google.com/scholar?q=Amita+Misra+and+Marilyn+Walker%2E+2013%2E+Topic+Independent+Identification+of+Agreement+and+Disagreement+in+Social+Media+Dialogue%2E+Proceedings+of+the+SIGDIAL+2013+Conference%2C+%28August%29%3A41%E2%80%9350%2E)
18. Radford M Neal. 2003. Slice sampling. The Annals of Statistics, 31(3):705–767. [find](https://scholar.google.com/scholar?q=Radford+M+Neal%2E+2003%2E+Slice+sampling%2E+The+Annals+of+Statistics%2C+31%283%29%3A705%E2%80%93767%2E)
19. James O’Shea, Zuhair Bandar, and Keeley Crockett. 2012. A Multi-classifier Approach to Dialogue Act Classification Using Function Words. Springer Berlin Heidelberg, Berlin, Heidelberg. [find](https://scholar.google.com/scholar?q=James+O%E2%80%99Shea%2C+Zuhair+Bandar%2C+and+Keeley+Crockett%2E+2012%2E+A+Multi%2Dclassifier+Approach+to+Dialogue+Act+Classification+Using+Function+Words%2E+Springer+Berlin+Heidelberg%2C+Berlin%2C+Heidelberg%2E)
20. Michael J. Paul. 2012. Mixed membership markov models for unsupervised conversation modeling. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 94–104, Jeju Island, Korea. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Michael+J%2E+Paul%2E+2012%2E+Mixed+membership+markov+models+for+unsupervised+conversation+modeling%2E+In+Proceedings+of+the+2012+Joint+Conference+on+Empirical+Methods+in+Natural+Language+Processing+and+Comput)
21. Alan Ritter, Colin Cherry, and Bill Dolan. 2010. [Unsupervised Modeling of Twitter Conversations](https://md.research.elamur.ai/free/en/papers/ritter-2010-unsupervised-modeling-twitter-conversations/). In Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pages 172–180, Los Angeles, California. Association for Computational Linguistics.
22. Andrew Rosenberg and Julia Hirschberg. 2007. [V-Measure: A Conditional Entropy-Based External Cluster Evaluation Measure](https://md.research.elamur.ai/free/en/papers/rosenberg-2007-vmeasure-conditional-entropy-cluster-evaluation/). In Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (EMNLP-CoNLL), pages 410–420, Prague, Czech Republic. Association for Computational Linguistics.
23. Jack Sidnell. 2011. Conversation Analysis: An Introduction. Language in Society. Wiley. [find](https://scholar.google.com/scholar?q=Jack+Sidnell%2E+2011%2E+Conversation+Analysis%3A+An+Introduction%2E+Language+in+Society%2E+Wiley%2E)
24. David R. Traum. 1999. Speech Acts for Dialogue Agents. Springer Netherlands, Dordrecht. [↗](https://doi.org/10.1007/978-94-015-9204-8_8)
25. Soroush Vosoughi and Deb Roy. 2016. [Tweet Acts : A Speech Act Classifier for Twitter](https://md.research.elamur.ai/free/en/papers/vosoughi-2016-tweet-acts-speech-act-classifier/). Proceedings of the 10th AAAI Conference on Weblogs and Social Media, (ICWSM):1–4. [↗](https://doi.org/10.48550/arxiv.1605.05156)
26. Terry Winograd and Fernando Flores. 1986. Understanding Computers and Cognition: A New Foundation for Design. Language and being. Ablex Publishing Corporation. [find](https://scholar.google.com/scholar?q=Terry+Winograd+and+Fernando+Flores%2E+1986%2E+Understanding+Computers+and+Cognition%3A+A+New+Foundation+for+Design%2E+Language+and+being%2E+Ablex+Publishing+Corporation%2E)
27. Britta Wrede and Elizabeth Shriberg. 2003. Relationship between dialogue acts and hot spots in meetings. In IEEE Workshop on Automatic Speech Recognition and Understanding, pages 180–185. IEEE. [find](https://scholar.google.com/scholar?q=Britta+Wrede+and+Elizabeth+Shriberg%2E+2003%2E+Relationship+between+dialogue+acts+and+hot+spots+in+meetings%2E+In+IEEE+Workshop+on+Automatic+Speech+Recognition+and+Understanding%2C+pages+180%E2%80%93185%2E+IEEE%2E)

*Updated 1 November 2026*