---
title: "Dialogue Act Classification in Group Chats with DAG-LSTMs"
description: "The Bloomberg authors propose DAG-LSTM, a conversation-level recurrent model… 87.69% accuracy. 39 references, cited by 2 papers here."
authors: ["Ozan Irsoy", "Rakesh Gosangi", "Haimin Zhang", "Mu-Hsin Wei", "Peter Lund", "Duccio Pappadopulo", "Brendan Fahy", "Neophytos Nephytou", "Camilo Ortiz"]
year: 2019
venue: "arXiv:1908.01821"
source: "https://arxiv.org/abs/1908.01821"
doi: "10.48550/arxiv.1908.01821"
arxiv: "1908.01821"
licence: "arXiv non-exclusive distribution licence (read only, no redistribution)"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/irsoy-2019-dialogue-act-classification-group-chats/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/irsoy-2019-dialogue-act-classification-group-chats/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2019](https://md.research.elamur.ai/free/en/papers/2019/page/6/) / Dialogue Act Classification in Group Chats with DAG-LSTMs

# Dialogue Act Classification in Group Chats with DAG-LSTMs

**Authors:** [Ozan Irsoy](https://md.research.elamur.ai/free/en/authors/ozan-irsoy/), Rakesh Gosangi, Haimin Zhang, Mu-Hsin Wei, Peter Lund, Duccio Pappadopulo, Brendan Fahy, Neophytos Nephytou and Camilo Ortiz · **Year:** 2019 · **Venue:** arXiv:1908.01821 · **Licence:** [arXiv non-exclusive distribution licence (read only, no redistribution)](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/1908.01821) · [PDF](https://arxiv.org/pdf/1908.01821) · [DOI 10.48550/arxiv.1908.01821](https://doi.org/10.48550/arxiv.1908.01821) · [arXiv 1908.01821](https://arxiv.org/abs/1908.01821) · [OpenAlex](https://openalex.org/W2964453985)

**Citations:** 2 citations (OpenAlex, 2026-10-14) · 2 in our library, of the 141,840 papers whose bibliographies we read

## Summary
The paper addresses dialogue act (DA) labelling in multi-party written chats. In these chats several threads interleave, so a fixed window of preceding messages often misses the relevant context. The Bloomberg authors propose DAG-LSTM, a conversation-level recurrent model derived from Tree-LSTMs. Each message node gets two predecessors: the immediately previous message and the previous message by the same participant. This lets the network build a per-speaker history while still following chat order. Merging several predecessors could make the memory cell grow out of control as paths multiply, so the cells are combined with an elementwise maximum instead of a sum. Each utterance is first encoded by a BiLSTM whose word states are max-pooled. On the STAC corpus of Settlers of Catan trade chats, BiLSTM + DAG-LSTM beat four other hierarchical CNN/LSTM combinations on both accuracy and macro-F1. Most of the gain came from fewer confusions between Offer and Counteroffer. The work shows that speaker-aware graph structure is a cheap and effective source of context for pragmatic labelling.

## Key points
- Data: the STAC corpus with five kept labels. Other has 7012 items, Offer 1571, Refusal 1546, Accept 658 and Counteroffer 643. Preference, with only 8 items, was dropped.
- Best model: BiLSTM + DAG-LSTM reached 87.69% accuracy, 0.86 points above BiLSTM + LSTM. Its macro-F1 of 75.78% was more than 1 point above the runner-up.
- The DAG-LSTM had the top F1 on four of the five classes. Its confusion matrix shows fewer Offer→Counteroffer and Refusal→Other errors.
- The context models were left unidirectional on purpose, because a bidirectional one cannot run on live chat.
- For reference, earlier CRF results from the STAC authors reported 83% accuracy and 73% macro-F1. They used a preliminary 10-game subset, so they are not directly comparable.

## Interesting ideas
- Give each turn an extra edge to the same speaker's previous turn, so a person's own earlier move can be reached in one hop even when other speakers interrupt.
- When a node has several incoming states, pick the elementwise maximum instead of summing them, which keeps magnitudes bounded and effectively follows one dominant path.
- Use only past context for the conversation model so the system works incrementally on live input.
- Judge the value of context by inspecting confusions between classes with near-identical surface wording (Offer vs Counteroffer), not only by the overall score.
- The authors suggest the same graph could also link turns by team or by thread, which turns context selection into a choice of edge types.

## For Mova

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

- **I1** (frames): speech-act tagging for dialogues in tales could treat "last turn overall" and "this speaker's last turn" as explicit antecedents. That would make explanations like "a counteroffer because this character offered before" traceable.
- **I2** (world): when an entity's state can be inherited from several earlier event branches, a select-one merge rule keeps each value tied to a single identifiable source, which preserves explainability.
- **I3** (answer): readers that process a story turn by turn could stay causal, so an answer never depends on text that comes later than the question point.
- **I4** (learn): evaluation reports could track specific confusable label pairs, for example speech acts that differ only by prior discourse. This would show whether a context feature actually helps where it should.
- **I5** (frames): coreference and implicature modules could consume typed links between turns (same speaker, same addressee, same topic) as separate, inspectable signals.

*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 (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 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: Ozan Irsoy, Rakesh Gosangi, Haimin Zhang, Mu-Hsin Wei, Peter Lund, Duccio Pappadopulo, Brendan Fahy, Neophytos Nephytou and Camilo Ortiz. 2019. Dialogue Act Classification in Group Chats with DAG-LSTMs. arXiv:1908.01821.

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

```bibtex
@misc{irsoy2019dialogue,
  title = {{Dialogue Act Classification in Group Chats with DAG-LSTMs}},
  author = {Ozan Irsoy, Rakesh Gosangi, Haimin Zhang, Mu-Hsin Wei, Peter Lund, Duccio Pappadopulo, Brendan Fahy, Neophytos Nephytou and Camilo Ortiz},
  year = {2019},
  doi = {10.48550/arxiv.1908.01821},
  eprint = {1908.01821},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1908.01821}
}
```

- [← previous Detecting Syntactic Change Using a Neural Part-of-Speech Tagger](https://md.research.elamur.ai/free/en/papers/merrill-2019-detecting-syntactic-change-neural-part/)
- [next → Experimental studies on it-clefts and predicate interpretation](https://md.research.elamur.ai/free/en/papers/renans-2019-experimental-studies-it-clefts-predicate/)
- [↑ back to papers of 2019](https://md.research.elamur.ai/free/en/papers/2019/page/6/)

## References

39 references, 15 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, and Elizabeth Shriberg. 2005. Automatic dialog act segmentation and classification in multiparty meetings. In Proceedings.(ICASSP’05). IEEE International Conference on Acoustics, Speech, and Signal Processing, 2005., Vol. 1. IEEE, I–1061. [find](https://scholar.google.com/scholar?q=Jeremy+Ang%2C+Yang+Liu%2C+and+Elizabeth+Shriberg%2E+2005%2E+Automatic+dialog+act+segmentation+and+classification+in+multiparty+meetings%2E+In+Proceedings%2E%28ICASSP%E2%80%9905%29%2E+IEEE+International+Conference+on+Acoustics%2C)
2. Nicholas Asher, Julie Hunter, Mathieu Morey, Farah Benamara, and Stergos D. Afantenos. 2016. [Discourse Structure and Dialogue Acts in Multiparty Dialogue: the STAC Corpus](https://md.research.elamur.ai/free/en/papers/asher-2016-discourse-structure-dialogue-acts-multiparty/). In LREC.
3. John Langshaw Austin. 1975. How to do things with words. Oxford university press. [↗](https://doi.org/10.1093/acprof:oso/9780198245537.001.0001)
4. Lina M Rojas Barahona, Milica Gasic, Nikola Mrkšić, Pei-Hao Su, Stefan Ultes, Tsung-Hsien Wen, and Steve Young. 2016. Exploiting sentence and context representations in deep neural models for spoken language understanding. arXiv preprint arXiv:1610.04120 (2016). Gold DAG-LSTM LSTM Offer Refusal Offer Counteroffer Refusal Offer Counteroffer Offer Refusal Offer Refusal Offer Refusal Offer Counteroffer Refusal Offer Counteroffer Offer Refusal Offer Refusal Offer Refusal Offer Counteroffer Refusal Counteroffer Counteroffer Counteroffer Refusal Counteroffer Refusal Offer Accept Other Counteroffer Other Other Offer Counteroffer Other Counteroffer Other Other Offer Offer Other Offer Refusal Other Offer Offer Offer Refusal Other Other Other Other Refusal Other Offer Offer Offer Refusal Other Other Other Other Other Other Offer Other Counteroffer Refusal Other Other Other Refusal Other Other [↗](https://arxiv.org/abs/1610.04120)
5. Harry Bunt, Jan Alexandersson, Jae-Woong Choe, Alex Chengyu Fang, Koiti Hasida, Volha Petukhova, Andrei Popescu-Belis, and David R Traum. 2012. [ISO 24617-2: A semantically-based standard for dialogue annotation](https://md.research.elamur.ai/free/en/papers/bunt-2012-iso-24617-2-semantically-based/).. In LREC. 430– 437.
6. Anais Cadilhac, Nicholas Asher, Farah Benamara, and Alex Lascarides. 2013. Grounding strategic conversation: Using negotiation dialogues to predict trades in a win-lose game. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing. 357–368. [find](https://scholar.google.com/scholar?q=Anais+Cadilhac%2C+Nicholas+Asher%2C+Farah+Benamara%2C+and+Alex+Lascarides%2E+2013%2E+Grounding+strategic+conversation%3A+Using+negotiation+dialogues+to+predict+trades+in+a+win%2Dlose+game%2E+In+Proceedings+of+the+201)
7. Qian Chen, Xiaodan Zhu, Zhenhua Ling, Si Wei, and Hui Jiang. 2016. Enhancing and combining sequential and tree lstm for natural language inference. arXiv preprint arXiv:1609.06038 (2016). [↗](https://arxiv.org/abs/1609.06038)
8. Xinyun Chen, Chang Liu, and Dawn Song. 2018. Tree-to-tree neural networks for program translation. In Advances in Neural Information Processing Systems. 2547–2557. [↗](https://doi.org/10.48550/arxiv.1802.03691)
9. Xinchi Chen, Zhan Shi, Xipeng Qiu, and Xuanjing Huang. 2017. DAG-based Long Short-Term Memory for Neural Word Segmentation. arXiv preprint arXiv:1707.00248 (2017). [↗](https://arxiv.org/abs/1707.00248)
10. Alfred Dielmann and Steve Renals. 2008. Recognition of dialogue acts in multi-party meetings using a switching DBN. IEEE transactions on audio, speech, and language processing 16, 7 (2008), 1303–1314. [find](https://scholar.google.com/scholar?q=Alfred+Dielmann+and+Steve+Renals%2E+2008%2E+Recognition+of+dialogue+acts+in+multi%2Dparty+meetings+using+a+switching+DBN%2E+IEEE+transactions+on+audio%2C+speech%2C+and+language+processing+16%2C+7+%282008%29%2C+1303%E2%80%931314%2E)
11. Raul Fernandez and Rosalind W Picard. 2002. Dialog act classification from prosodic features using support vector machines. In Speech Prosody 2002, International Conference. [find](https://scholar.google.com/scholar?q=Raul+Fernandez+and+Rosalind+W+Picard%2E+2002%2E+Dialog+act+classification+from+prosodic+features+using+support+vector+machines%2E+In+Speech+Prosody+2002%2C+International+Conference%2E)
12. Eric N Forsythand and Craig H Martell. 2007. Lexical and discourse analysis of online chat dialog. In International Conference on Semantic Computing (ICSC 2007). IEEE, 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%2E+IEEE%2C+19%E2%80%9326%2E)
13. Alex Graves and Jürgen Schmidhuber. 2005. Framewise phoneme classification with bidirectional LSTM and other neural network architectures. Neural Networks 18, 5-6 (2005), 602–610. [↗](https://doi.org/10.1016/j.neunet.2005.06.042)
14. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural computation 9, 8 (1997), 1735–1780. [↗](https://doi.org/10.1162/neco.1997.9.8.1735)
15. Daniel Jurafsky, Rebecca Bates, Noah Coccaro, Rachel Martin, Marie Meteer, Klaus Ries, Elizabeth Shriberg, Andreas Stolcke, Paul Taylor, and Carol Van Ess-Dykema. 1997. Automatic detection of discourse structure for speech recognition and understanding. In 1997 IEEE Workshop on Automatic Speech Recognition and Understanding Proceedings. IEEE, 88–95. [↗](https://doi.org/10.5445/ir/1000166547)
16. Nal Kalchbrenner and Phil Blunsom. 2013. [Recurrent Convolutional Neural Networks for Discourse Compositionality](https://md.research.elamur.ai/free/en/papers/kalchbrenner-2013-recurrent-convolutional-neural-networks-discourse/). In Proceedings of the Workshop on Continuous Vector Space Models and their Compositionality. 119–126.
17. Su Nam Kim, Lawrence Cavedon, and Timothy Baldwin. 2010. [Classifying dialogue acts in one-on-one live chats](https://md.research.elamur.ai/free/en/papers/kim-2010-classifying-dialogue-acts-one-one/). In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 862–871.
18. Su Nam Kim, Lawrence Cavedon, and Timothy Baldwin. 2012. [Classifying dialogue acts in multi-party live chats](https://md.research.elamur.ai/free/en/papers/kim-2012-classifying-dialogue-acts-multi-party/). In Proceedings of the 26th Pacific Asia Conference on Language, Information, and Computation. 463–472.
19. Diederik P Kingma and Jimmy Ba. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014). [↗](https://arxiv.org/abs/1412.6980)
20. Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao, and Shuzi Niu. 2017. [DailyDialog: A Manually Labelled Multi-turn Dialogue Dataset](https://md.research.elamur.ai/free/en/papers/li-2017-dailydialog-manually-labelled-multi-turn/). In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 986–995. [↗](https://doi.org/10.48550/arxiv.1710.03957)
21. Yang Liu, Kun Han, Zhao Tan, and Yun Lei. 2017. [Using context information for dialog act classification in DNN framework](https://md.research.elamur.ai/free/en/papers/liu-2017-context-information-dialog-act-classification/). In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. 2170–2178.
22. Jean Maillard, Stephen Clark, and Dani Yogatama. 2017. Jointly learning sentence embeddings and syntax with unsupervised tree-lstms. arXiv preprint arXiv:1705.09189 (2017). [↗](https://arxiv.org/abs/1705.09189)
23. Christopher Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven Bethard, and David McClosky. 2014. [The Stanford CoreNLP natural language processing toolkit](https://md.research.elamur.ai/free/en/papers/manning-2014-stanford-corenlp-natural-language-processing/). In Proceedings of 52nd annual meeting of the association for computational linguistics: system demonstrations. 55–60.
24. Michael McTear, Zoraida Callejas, and David Griol. 2016. The conversational interface: Talking to smart devices. Springer. [↗](https://doi.org/10.1007/978-3-319-32967-3)
25. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. [Glove: Global vectors for word representation](https://md.research.elamur.ai/free/en/papers/pennington-2014-glove-global-vectors-word-representation/). In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP). 1532–1543.
26. Eugénio Ribeiro, Ricardo Ribeiro, and David Martins de Matos. 2015. [The influence of context on dialogue act recognition](https://md.research.elamur.ai/free/en/papers/ribeiro-2015-influence-context-dialogue-act-recognition/). arXiv preprint arXiv:1506.00839 (2015). [↗](https://arxiv.org/abs/1506.00839)
27. Eugénio Ribeiro, Ricardo Ribeiro, and David Martins de Matos. 2018. [Deep Dialog Act Recognition using Multiple Token, Segment, and Context Information Representations](https://md.research.elamur.ai/free/en/papers/ribeiro-2018-deep-dialog-act-recognition-using/). arXiv preprint arXiv:1807.08587 (2018). [↗](https://arxiv.org/abs/1807.08587)
28. John R Searle and John Rogers Searle. 1969. [Speech acts](https://md.research.elamur.ai/free/en/papers/searle-1969-speech-acts/): An essay in the philosophy of language. Vol. 626. Cambridge university press. [↗](https://doi.org/10.1017/cbo9781139173438)
29. Iulian Vlad Serban, Ryan Lowe, Peter Henderson, Laurent Charlin, and Joelle Pineau. 2018. A survey of available corpora for building data-driven dialogue systems: The journal version. Dialogue & Discourse 9, 1 (2018), 1–49. [↗](https://doi.org/10.5087/dad.2018.101)
30. 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/). In Proceedings of the 5th SIGdial Workshop on Discourse and Dialogue at HLT-NAACL 2004. [↗](https://doi.org/10.21236/ada460980)
31. Vivek Kumar Rangarajan Sridhar, Srinivas Bangalore, and Shrikanth Narayanan. 2009. Combining lexical, syntactic and prosodic cues for improved online dialog act tagging. Computer Speech & Language 23, 4 (2009), 407–422. [↗](https://doi.org/10.1016/j.csl.2008.12.001)
32. Andreas Stolcke, Klaus Ries, Noah Coccaro, Elizabeth Shriberg, Rebecca Bates, Daniel Jurafsky, Paul Taylor, Rachel Martin, Carol Van 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, 3 (2000), 339–373. [↗](https://doi.org/10.1162/089120100561737)
33. Dinoj Surendran and Gina-Anne Levow. 2006. Dialog act tagging with support vector machines and hidden Markov models. In Ninth International Conference on Spoken Language Processing. [find](https://scholar.google.com/scholar?q=Dinoj+Surendran+and+Gina%2DAnne+Levow%2E+2006%2E+Dialog+act+tagging+with+support+vector+machines+and+hidden+Markov+models%2E+In+Ninth+International+Conference+on+Spoken+Language+Processing%2E)
34. Kai Sheng Tai, Richard Socher, and Christopher D Manning. 2015. Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), Vol. 1. 1556–1566. [find](https://scholar.google.com/scholar?q=Kai+Sheng+Tai%2C+Richard+Socher%2C+and+Christopher+D+Manning%2E+2015%2E+Improved+Semantic+Representations+From+Tree%2DStructured+Long+Short%2DTerm+Memory+Networks%2E+In+Proceedings+of+the+53rd+Annual+Meeting+of+the)
35. Tsung-Hsien Wen, Milica Gasic, Nikola Mrkšić, Pei-Hao Su, David Vandyke, and Steve Young. 2015. [Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems](https://md.research.elamur.ai/free/en/papers/wen-2015-semantically-conditioned-lstm-based-natural/). In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing. 1711–1721.
36. Steve Young, Milica Gašić, Blaise Thomson, and Jason D Williams. 2013. Pomdp-based statistical spoken dialog systems: A review. Proc. IEEE 101, 5 (2013), 1160–1179. [↗](https://doi.org/10.1109/jproc.2012.2225812)
37. Tiancheng Zhao and Maxine Eskenazi. 2016. Towards End-to-End Learning for Dialog State Tracking and Management using Deep Reinforcement Learning. In 17th Annual Meeting of the Special Interest Group on Discourse and Dialogue. 1. [find](https://scholar.google.com/scholar?q=Tiancheng+Zhao+and+Maxine+Eskenazi%2E+2016%2E+Towards+End%2Dto%2DEnd+Learning+for+Dialog+State+Tracking+and+Management+using+Deep+Reinforcement+Learning%2E+In+17th+Annual+Meeting+of+the+Special+Interest+Group+o)
38. Xiaodan Zhu, Parinaz Sobhani, and Hongyu Guo. 2016. Dag-structured long short-term memory for semantic compositionality. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 917–926. [find](https://scholar.google.com/scholar?q=Xiaodan+Zhu%2C+Parinaz+Sobhani%2C+and+Hongyu+Guo%2E+2016%2E+Dag%2Dstructured+long+short%2Dterm+memory+for+semantic+compositionality%2E+In+Proceedings+of+the+2016+Conference+of+the+North+American+Chapter+of+the+Asso)
39. Xiaodan Zhu, Parinaz Sobihani, and Hongyu Guo. 2015. Long short-term memory over recursive structures. In International Conference on Machine Learning. 1604– 1612. [find](https://scholar.google.com/scholar?q=Xiaodan+Zhu%2C+Parinaz+Sobihani%2C+and+Hongyu+Guo%2E+2015%2E+Long+short%2Dterm+memory+over+recursive+structures%2E+In+International+Conference+on+Machine+Learning%2E+1604%E2%80%93+1612%2E)

*Updated 24 October 2026*