---
title: "Dialogue Act Recognition using Reweighted Speaker Adaptation"
description: "The paper asks whether dialogue act recognition in multiparty meetings improves… 76.76% accuracy. 16 references, cited by 2 papers here."
authors: ["Congkai Sun", "Louis-Philippe Morency"]
year: 2012
venue: "Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue"
source: "https://aclanthology.org/W12-1616/"
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/sun-2012-dialogue-act-recognition-reweighted-speaker/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/sun-2012-dialogue-act-recognition-reweighted-speaker/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2012](https://md.research.elamur.ai/free/en/papers/2012/page/2/) / Dialogue Act Recognition using Reweighted Speaker Adaptation

# Dialogue Act Recognition using Reweighted Speaker Adaptation

**Authors:** Congkai Sun and [Louis-Philippe Morency](https://md.research.elamur.ai/free/en/authors/louis-philippe-morency/) · **Year:** 2012 · **Venue:** [Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue](https://md.research.elamur.ai/free/en/venues/sigdial/) ([SIGDIAL 2012](https://md.research.elamur.ai/free/en/venues/sigdial/2012/)) · **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/W12-1616/) · [PDF](https://aclanthology.org/W12-1616.pdf) · [OpenAlex](https://openalex.org/W2118564791)

**Citations:** 4 citations (OpenAlex, 2026-10-14) · [2](https://md.research.elamur.ai/free/en/papers/sun-2012-dialogue-act-recognition-reweighted-speaker/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
The paper asks whether dialogue act recognition in multiparty meetings improves when the model adapts to the individual speaker. A single generic classifier struggles because people phrase the same intent differently and sometimes mean different things with identical words. The authors frame each speaker as a separate domain, which turns the task into supervised domain adaptation. Using seven frequent speakers from the ICSI-MRDA corpus, five general tags and MaxEnt classifiers over n-gram features, they first check that same-speaker training beats other-speaker training. They then study qualitatively how speakers differ. Their method pools all other speakers' data with a small speaker-specific set and multiplies the weight of the speaker's own examples by a factor tuned on held-out data from that speaker. This reweighted adaptation beats a generic model, a speaker-only model and plain data pooling, even with only 200 annotated acts from the target speaker. The result matters for settings like recurring meetings, where archives of past talk by the same people make a little per-person annotation affordable.

## Key points
- Data: the ICSI-MRDA corpus (75 meetings, 53 unique speakers). Seven speakers with more than 4,000 dialogue acts each were selected, giving 47,040 dialogue acts. The speaker me013 was split by meeting type (Bmr vs Bro).
- Tags: the 5 general MRDA tags are Disruption, Back Channel, Floor Mechanism, Question and Statement. Statement dominates at about 55% of the data.
- Features and model: MaxEnt classifiers over unigrams (14,653), bigrams (158,884) and trigrams (400,025) extracted from transcriptions.
- The generic recognizer, trained on all other speakers, reaches 76.76% accuracy. A recognizer trained only on 2,000 speaker-specific acts reaches just 71.06%.
- Speaker data was varied over 200, 500, 1000, 1500 and 2000 acts. Candidate weights were 10, 30, 50, 75 and 100, chosen by validation. The best weight decreases as more speaker-specific data becomes available.

## Interesting ideas
- Before building an adaptation method, test whether the domains really differ: train on one speaker and compare scores on the same speaker versus other speakers. This guards against adapting to noise.
- Pooling a small in-domain set with a large out-of-domain set by plain concatenation drowns the small set. Upweighting the small set's examples by a tuned constant is a nearly free fix.
- The weight should be chosen on held-out data from the target domain, and the best weight naturally shrinks as in-domain data grows. This gives a principled schedule instead of a fixed constant.
- Measuring mutual information between an exact utterance and act labels, separately per speaker, exposes "conflicts" where the same words carry different acts. The word "right" is one example.
- A small amount of speaker-specific data alone loses to a generic model, yet the same small amount added on top of the generic data wins. The value of personalization lies in combining it with generic data.

## For Mova

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

- **I1** (learn): Before training genre- or author-specific variants of parsers or taggers, run the same "train here, test here vs there" check as a cheap control.
- **I2** (learn): When mixing a small gold treebank with large silver data from witness parsers, give gold sentences a weight factor rather than treating all examples equally.
- **I3** (learn): Tune the gold-vs-silver weight on a gold dev split and re-tune it as the gold set grows.
- **I4** (frames): Use per-source label statistics to flag cue words whose speech-act reading depends on context, so the prag crate resolves them from context rather than from a fixed lexical rule.
- **I5** (frames): Keep a generic speech-act model and layer lightweight per-text or per-character adjustments on top, for example the habits of a recurring character in a tale.

*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

- [USC Institute for Creative Technologies](https://md.research.elamur.ai/free/en/labs/usc-institute-for-creative-technologies/) — Research institute, Playa Vista, United States · 4 works · 278 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: Congkai Sun and Louis-Philippe Morency. 2012. Dialogue Act Recognition using Reweighted Speaker Adaptation. Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue.

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

```bibtex
@inproceedings{sun2012dialogue,
  title = {{Dialogue Act Recognition using Reweighted Speaker Adaptation}},
  author = {Congkai Sun and Louis-Philippe Morency},
  booktitle = {Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue},
  year = {2012},
  url = {https://aclanthology.org/W12-1616/}
}
```

- [← previous Diachronic Changes in Text Complexity in 20th Century English Language: An NLP Approach](https://md.research.elamur.ai/free/en/papers/stajner-2012-diachronic-changes-text-complexity-20th/)
- [next → Exploiting Discourse Relations for Sentiment Analysis](https://md.research.elamur.ai/free/en/papers/wang-2012-exploiting-discourse-relations-sentiment-analysis/)
- [↑ back to papers of 2012](https://md.research.elamur.ai/free/en/papers/2012/page/2/)

## References

16 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Jeremy Ang, Yang Liu, Elizabeth Shriberg 2005. Automatic Dialog Act Segmentation and Classification in Multiparty Meetings. ICASSP. [find](https://scholar.google.com/scholar?q=Jeremy+Ang%2C+Yang+Liu%2C+Elizabeth+Shriberg+2005%2E+Automatic+Dialog+Act+Segmentation+and+Classification+in+Multiparty+Meetings%2E+ICASSP%2E)
2. Jean Carletta. 2007. Unleashing the killer corpus: experiences in creating the multi-everything AMI Meeting Corpus. Language Resources and Evaluation, 41(2): 181-190 [↗](https://doi.org/10.1007/s10579-007-9040-x)
3. Mark Core and James Allen. 1997. Working Notes: AAAI Fall Symposium. HLT-NAACL SIGDIAL Workshop. [find](https://scholar.google.com/scholar?q=Mark+Core+and+James+Allen%2E+1997%2E+Working+Notes%3A+AAAI+Fall+Symposium%2E+HLT%2DNAACL+SIGDIAL+Workshop%2E)
4. Hal Daumé III. 2007. [Frustratingly Easy Domain Adaptation](https://md.research.elamur.ai/free/en/papers/iii-2007-frustratingly-easy-domain-adaptation/). Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics.
5. Raquel Fernandez, Matthew Frampton, Patrick Ehlen, Matthew Purver and Stanley Peters. 2008. Modelling and Detecting Decisions in Multi-Party Dialogue. Proceedings of the 9th SIGdial Workshop on Discourse and Dialogue. [find](https://scholar.google.com/scholar?q=Raquel+Fernandez%2C+Matthew+Frampton%2C+Patrick+Ehlen%2C+Matthew+Purver+and+Stanley+Peters%2E+2008%2E+Modelling+and+Detecting+Decisions+in+Multi%2DParty+Dialogue%2E+Proceedings+of+the+9th+SIGdial+Workshop+on+Discou)
6. Umit Guz, Gokhan Tur, Dilek Hakkani-Tur, and Sebastien Cuendet. 2009. Cascaded model adaptation for dialog act segmentation and tagging. Computer Speech & Language, 24(2):289–306. [↗](https://doi.org/10.1016/j.csl.2009.04.006)
7. Minwoo Jeong, Chin-Yew Lin and Gary Lee. 2009. Semi-supervised speech act recognition in emails and forums. The 2009 Conference on Empirical Methods on Natural Language Processing. [find](https://scholar.google.com/scholar?q=Minwoo+Jeong%2C+Chin%2DYew+Lin+and+Gary+Lee%2E+2009%2E+Semi%2Dsupervised+speech+act+recognition+in+emails+and+forums%2E+The+2009+Conference+on+Empirical+Methods+on+Natural+Language+Processing%2E)
8. Jing Jiang and ChengXiang Zhai. 2007. [Instance weighting for domain adaptation in NLP](https://md.research.elamur.ai/free/en/papers/jiang-2007-instance-weighting-domain-adaptation-nlp/). Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics.
9. Jachym Kolar, Yang Liu, and Elizabeth Shriberg. 2007. Speaker Adaptation of Language Models for Automatic Dialog Act Segmentation of Meetings. Interspeech, 339–373. [↗](https://doi.org/10.21236/ada469307)
10. Anna Margolis, Karen Livescu, Mari Ostendorf. 2010. Semi-supervised domain adaptation for automatic dialog act tagging. ACL 2010 Workshop on Domain Adaptation for Natural Language Processing. [find](https://scholar.google.com/scholar?q=Anna+Margolis%2C+Karen+Livescu%2C+Mari+Ostendorf%2E+2010%2E+Semi%2Dsupervised+domain+adaptation+for+automatic+dialog+act+tagging%2E+ACL+2010+Workshop+on+Domain+Adaptation+for+Natural+Language+Processing%2E)
11. Rajesh Ranganath, Dan Jurafsky, and Dan McFarland. 2009. It’s Not You, it’s Me: Detecting Flirting and its Misperception in Speed-Dates. The 2009 Conference on Empirical Methods on Natural Language Processing. [find](https://scholar.google.com/scholar?q=Rajesh+Ranganath%2C+Dan+Jurafsky%2C+and+Dan+McFarland%2E+2009%2E+It%E2%80%99s+Not+You%2C+it%E2%80%99s+Me%3A+Detecting+Flirting+and+its+Misperception+in+Speed%2DDates%2E+The+2009+Conference+on+Empirical+Methods+on+Natural+Language+Pr)
12. Vivek Rangarajan, Srinivas Bangaloreb and Shrikanth Narayanana. 2009. Combining lexical, syntactic and prosodic cues for improved online dialog act tagging. Computer Speech and Language, 23(4): 407-422 [find](https://scholar.google.com/scholar?q=Vivek+Rangarajan%2C+Srinivas+Bangaloreb+and+Shrikanth+Narayanana%2E+2009%2E+Combining+lexical%2C+syntactic+and+prosodic+cues+for+improved+online+dialog+act+tagging%2E+Computer+Speech+and+Language%2C+23%284%29%3A+407%2D42)
13. Adwait Ratnaparkhi. 1996. [A maximum entropy model for part-of-speech tagging](https://md.research.elamur.ai/free/en/papers/ratnaparkhi-1996-maximum-entropy-model-part-speech/). In Proceedings of the Conference on Empirical Methods in Natural Language Processing.
14. Elizabeth Shriberg, Raj Dhillon, Sonali Bhagat, Jeremy Ang and Hannah Carvey. 2004. [The ICSI Meeting Recorder Dialog Act (MRDA) Corpus](https://md.research.elamur.ai/free/en/papers/shriberg-2004-icsi-meeting-recorder-dialog-act/). HLT-NAACL SIGDIAL Workshop. [↗](https://doi.org/10.21236/ada460980)
15. Andreas Stolcke, Klaus Ries, Noah Coccaro, Elizabeth Shriberg, Rebecca Bates, Daniel Jurafsky, Paul Taylor, Rachel Martin, Carol V. Ess-dykema and Marie Meteer. 2000. [Dialogue Act Modeling for Automatic Tagging and Recognition of Conversational Speech](https://md.research.elamur.ai/free/en/papers/stolcke-2000-dialogue-act-modeling-automatic-tagging/). Computational Linguistics, 26:339-373. [↗](https://doi.org/10.1162/089120100561737)
16. Gokhan Tur, Umit Guz and Dilek Hakkani-Tur. 2006. Model Adaptation For Dialogue Act Tagging. Spoken Language Technology Workshop. [find](https://scholar.google.com/scholar?q=Gokhan+Tur%2C+Umit+Guz+and+Dilek+Hakkani%2DTur%2E+2006%2E+Model+Adaptation+For+Dialogue+Act+Tagging%2E+Spoken+Language+Technology+Workshop%2E)

*Updated 24 October 2026*