Elamur Research
☰
home / papers / 2017 / Dialogue Act Sequence Labeling using Hierarchical encoder with CRF

Dialogue Act Sequence Labeling using Hierarchical encoder with CRF

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive licence to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/1709.04250

Summary

Dialogue act (DA) recognition assigns a semantic tag, such as Greeting, Statement or Y/N question, to each utterance in a conversation. Earlier systems had two weaknesses. Some depended on hand-engineered features that transfer poorly between corpora. Others modelled only one kind of dependency: either between neighbouring utterances or between neighbouring labels.

Kumar et al. (IBM Research India) treat the task as sequence labeling and build a hierarchical Bi-LSTM-CRF that is trained end-to-end:

  • a word-level Bi-LSTM compresses each utterance into a vector;
  • a second Bi-LSTM runs over these vectors to encode conversational context;
  • a linear-chain CRF on top decodes the whole label sequence jointly, so links among utterances and among tags are both modelled.

The model was tested on the Switchboard (SwDA) and Meeting Recorder (MRDA) benchmarks. It outperforms prior state of the art by 2.2 and 4.1 absolute points respectively, reaching roughly 79% on SwDA. Human inter-annotator agreement on that corpus is 84%. The results show that using the nested structure of dialogue together with label structure pays off, which is useful for dialogue systems and conversation summarizers.

Key points

  • Accuracy: 79.2% on SwDA (previous best 77%) and 90.9% on MRDA (previous best 86.8% from LSTM-Softmax by Khanpour et al.; CNN by Lee and Dernoncourt reached 84.6%).
  • Ablation on SwDA:
    • word embeddings alone give 72.2% with CRF and 71.4% with logistic regression;
    • adding the utterance and conversation layers raises this to 79.2% with CRF but only 74.1% with LR;
    • the conversation layer and the CRF output are therefore the decisive parts.
  • Adding a parallel POS-tag encoder lowered accuracy to 77.9%, about 1 point.
  • Intra-attention context over the previous K utterances gave 77.4%, 78.3% and 78.1% for K = 10, 5 and 3. None of these beat the base model, which suggests the current DA depends mostly on the last 2–3 turns.
  • Attention context did speed up training. After the first epoch it reached 68.8% vs 65.1% on SwDA and 88% vs 87% on MRDA.

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

Fields: Linguistics & languages

Topics: Speech act theory

Cited by

48 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 44 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1709.04250 · arXiv 1709.04250

@misc{kumar2017dialogue,
  title = {{Dialogue Act Sequence Labeling using Hierarchical encoder with CRF}},
  author = {Kumar, Harshit and Agarwal, Arvind and Dasgupta, Riddhiman and Joshi, Sachindra and Kumar, Arun},
  year = {2017},
  doi = {10.48550/arxiv.1709.04250},
  eprint = {1709.04250},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1709.04250}
}

References

33 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ang, J.; Liu, Y.; and Shriberg, E. 2005. Automatic dialog act segmentation and classification in multiparty meetings. In ICASSP. find
  2. Bahdanau, D.; Cho, K.; and Bengio, Y. 2015. Neural machine translation by jointly learning to align and translate. ICLR. find our card
  3. Blunsom, P., and Kalchbrenner, N. 2013. Recurrent convolutional neural networks for discourse compositionality. In Proceedings of the 2013 Workshop on Continuous Vector Space Models and their Compositionality. find our card
  4. Cho, K.; van Merriënboer, B.; Bahdanau, D.; and Bengio, Y. 2014. On the properties of neural machine translation: Encoder–decoder approaches. Syntax, Semantics and Structure in Statistical Translation. find our card
  5. Core, M. G., and Allen, J. 1997. Coding dialogs with the damsl annotation scheme. In AAAI Fall Symposium On Communicative Action In Humans And Machines. find
  6. Dehghani, M.; Rothe, S.; Alfonseca, E.; and Fleury, P. 2017. Learning to attend, copy, and generate for session based query suggestion. In CIKM. find
  7. Grau, S.; Sanchis, E.; Castro, M. J.; and Vilar, D. 2004. Dialogue act classification using a bayesian approach. In 9th Conference Speech and Computer. find
  8. Higashinaka, R.; Imamura, K.; Meguro, T.; Miyazaki, C.; Kobayashi, N.; Sugiyama, H.; Hirano, T.; Makino, T.; and Matsuo, Y. 2014. Towards an open-domain conversational system fully based on natural language processing. In COLING. find
  9. Hochreiter, S., and Schmidhuber, J. 1997. Long short-term memory. Neural computation 9. find
  10. Huang, Z.; Xu, W.; and Yu, K. 2015. Bidirectional lstm-crf models for sequence tagging. arXiv preprint arXiv:1508.01991. ↗ our card
  11. Janin, A.; Baron, D.; Edwards, J.; Ellis, D.; Gelbart, D.; Morgan, N.; Peskin, B.; Pfau, T.; Shriberg, E.; Stolcke, A.; et al. 2003. The icsi meeting corpus. In ICASSP. find
  12. Ji, Y.; Haffari, G.; and Eisenstein, J. 2016. A latent variable recurrent neural network for discourse relation language models. In NAACL-HLT. find
  13. Jurafsky, D. 1997. Switchboard swbd-damsl shallow-discourse-function annotation coders manual. www. dcs. shef. ac. uk/nlp/amities/files/bib/ics-tr-97-02. pdf. find
  14. Khanpour, H.; Guntakandla, N.; and Nielsen, R. 2016. Dialogue act classification in domain-independent conversations using a deep recurrent neural network. In COLING. find our card
  15. Kim, Y.; Jernite, Y.; Sontag, D.; and Rush, A. M. 2016. Character-aware neural language models. In AAAI. find our card
  16. Lafferty, J. D.; McCallum, A.; and Pereira, F. C. N. 2001. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In ICML. find
  17. LeCun, Y.; Bengio, Y.; and Hinton, G. 2015. Deep learning. Nature. find
  18. Lee, J. Y., and Dernoncourt, F. 2016. Sequential short-text classification with recurrent and convolutional neural networks. In Proceedings of NAACL-HLT. find our card
  19. Lendvai, P., and Geertzen, J. 2007. Token-based chunking of turn-internal dialogue act sequences. In SIGDIAL Workshop on Discourse and Dialogue. find
  20. Ma, X., and Hovy, E. 2016. End-to-end sequence labeling via bi-directional lstm-cnns-crf. In ACL. find our card
  21. Mikolov, T.; Sutskever, I.; Chen, K.; Corrado, G. S.; and Dean, J. 2013. Distributed representations of words and phrases and their compositionality. In NIPS. find
  22. Paulus, R.; Xiong, C.; and Socher, R. 2017. A deep reinforced model for abstractive summarization. arXiv preprint arXiv:1705.04304. ↗ our card
  23. Pennington, J.; Socher, R.; and Manning, C. D. 2014. Glove: Global vectors for word representation. In EMNLP. find
  24. Rabiner, L. R. 1989. A tutorial on hidden markov models and selected applications in speech recognition. Proceedings of the IEEE. find
  25. Ribeiro, E.; Ribeiro, R.; and de Matos, D. M. 2015. The influence of context on dialogue act recognition. arXiv preprint arXiv:1506.00839. ↗
  26. Sennrich, R.; Haddow, B.; and Birch, A. 2017. Neural machine translation of rare words with subword units. In ACL. find
  27. Serban, I. V.; Sordoni, A.; Bengio, Y.; Courville, A. C.; and Pineau, J. 2016. Building end-to-end dialogue systems using generative hierarchical neural network models. In AAAI. find our card
  28. Serban, I. V.; Sordoni, A.; Lowe, R.; Charlin, L.; Pineau, J.; Courville, A. C.; and Bengio, Y. 2017. A hierarchical latent variable encoder-decoder model for generating dialogues. In AAAI. find our card
  29. Sordoni, A.; Bengio, Y.; Vahabi, H.; Lioma, C.; Grue Simonsen, J.; and Nie, J.-Y. 2015. A hierarchical recurrent encoder-decoder for generative context-aware query suggestion. In CIKM. find
  30. Stolcke, A.; Ries, K.; Coccaro, N.; Shriberg, E.; Bates, R.; Jurafsky, D.; Taylor, P.; Martin, R.; Van Ess-Dykema, C.; and Meteer, M. 2006. Dialogue act modeling for automatic tagging and recognition of conversational speech. Dialogue 26(3). find our card
  31. Tavafi, M.; Mehdad, Y.; Joty, S. R.; Carenini, G.; and Ng, R. T. 2013. Dialogue act recognition in synchronous and asynchronous conversations. In SIGDIAL. find
  32. Viterbi, A. 1967. Error bounds for convolutional codes and an asymptotically optimum decoding algorithm. IEEE Transactions on Information Theory. find
  33. Zimmermann, M. 2009. Joint segmentation and classification of dialog acts using conditional random fields. In InterSpeech. find