Elamur Research⌕
☰
home / papers / 2015 / Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems

Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems

Summary

Spoken dialogue systems usually produce responses from hand-written rules or templates. These are reliable but sound repetitive and are costly to extend to new domains or languages. Wen et al. propose SC-LSTM, a statistical generator that adds a sentence-planning component to a standard LSTM language model. A control vector represents which kind of dialogue act (DA) is being expressed, together with the slot-value pairs attached to it. A learned sigmoid gate gradually removes information from this vector as the corresponding content is expressed. The model trains on delexicalised, unaligned utterances with ordinary cross entropy, so sentence planning and surface realisation are learned jointly without the heuristic gating of the authors' earlier RNN generator. Varied outputs come from sampling candidates, which a backward model then reranks. Experiments covered two San Francisco venue domains, restaurants and hotels. SC-LSTM, especially its deep variant, scored best on objective metrics. Human judges also rated it more informative and natural than class-based LM and heuristic-gate RNN baselines. The result matters because it shows that data-driven, end-to-end NLG can replace handcrafting while still conveying the required meaning.

Key points

  • SC-LSTM cell: a conventional LSTM handles surface realisation, while a separate cell holds a 1-hot DA/slot-value vector d. A learned sigmoid gate controls this vector, so no exact-match heuristics are needed.
  • Prior work's limitation: the earlier Wen et al. (2015) system switched off slot features by exact string matching. This failed for binary slots and "dontcare" values; SC-LSTM learns this control instead.
  • Setup and baselines: evaluation used two domains (SF restaurant, SF hotel) and compared classlm, rnn w/ (heuristic gates), sc-lstm and +deep.
  • Human evaluation: judges rated 1000 DAs, about 1300 utterances per system after identical outputs were removed. SC-LSTM variants won on informativeness and naturalness and were significantly preferred in pairwise tests.
  • Deep vs. shallow: +deep was preferred most often, but its advantage over the shallow sc-lstm was not statistically significant.

Interesting ideas

  • I1. Meaning as a depletable budget: the content to express is stored as a vector that the model "uses up" as it generates. This idea could apply to any conditional generation task where omitting or repeating content is the main error, such as data-to-text or report generation.
  • I2. Delexicalisation: slot values are replaced with placeholder tokens during training and filled back in afterwards. This shrinks the vocabulary and lets the model generalise to unseen values, which is useful for low-resource languages and domain transfer.
  • I3. Overgenerate, then rerank: sampling several candidates and rescoring them with a backward model gives cheap fluency gains and natural variation without decoding tricks.
  • I4. Inspecting learned gates: plotting gate activations shows which words realise which slot (Figure 3). This is a simple interpretability check that can be reused for other gated conditioning mechanisms.
  • I5. Fairer pairwise judging: outputs that both systems produce identically are removed before human comparison. This focuses judges' effort on the cases where the systems actually differ.

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: Natural language processing

Cited by

318 papers of our library cite it (of the 140884 whose bibliographies we read); 21 of them have a page here, 297 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/d15-1199

@inproceedings{wen2015semantically,
  title = {{Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems}},
  author = {Wen, Tsung-Hsien and Gašić, Milica and Mrkšić, Nikola and Su, Pei-Hao and Vandyke, David and Young, Steve},
  booktitle = {Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing},
  year = {2015},
  doi = {10.18653/v1/d15-1199},
  url = {https://doi.org/10.18653/v1/d15-1199}
}

References

52 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Gabor Angeli, Percy Liang, and Dan Klein. 2010. A simple domain-independent probabilistic approach to generation. In Proceedings of the 2010 Conference on EMNLP, EMNLP ’10. Association for Computational Linguistics. find our card
  2. Frédéric Bastien, Pascal Lamblin, Razvan Pascanu, James Bergstra, Ian J. Goodfellow, Arnaud Bergeron, Nicolas Bouchard, and Yoshua Bengio. 2012. Theano: new features and speed improvements. find
  3. Deep Learning and Unsupervised Feature Learning NIPS 2012 Workshop. find
  4. Anja Belz. 2008. Automatic generation of weather forecast texts using comprehensive probabilistic generation-space models. Natural Language Engineering. find
  5. Yoshua Bengio, Patrice Simard, and Paolo Frasconi. 1994. Learning long-term dependencies with gradient descent is difficult. Neural Networks, IEEE Transactions on. find
  6. James Bergstra, Olivier Breuleux, Frédéric Bastien, Pascal Lamblin, Razvan Pascanu, Guillaume Desjardins, Joseph Turian, David Warde-Farley, and Yoshua Bengio. 2010. Theano: a CPU and GPU math expression compiler. In Proceedings of the Python for Scientific Computing Conference. find
  7. Adam Cheyer and Didier Guzzoni. 2007. Method and apparatus for building an intelligent automated assistant. US Patent App. 11/518,292. find
  8. Ronan Collobert and Jason Weston. 2008. A unified architecture for natural language processing: Deep neural networks with multitask learning. In Proceedings of the 25th International Conference on Machine Learning. find
  9. Milica Gašić, Dongho Kim, Pirros Tsiakoulis, Catherine Breslin, Matthew Henderson, Martin Szummer, Blaise Thomson, and Steve Young. 2014. Incremental on-line adaptation of pomdp-based dialogue managers to extended domains. In In Proceedings on InterSpeech. find
  10. Milica Gašić, Dongho Kim, Pirros Tsiakoulis, and Steve Young. 2015. Distributed dialogue policies for multi-domain statistical dialogue management. In In Proceedings on ICASSP. find
  11. Alex Graves, Marcus Liwicki, Santiago Fernández, Roman Bertolami, Horst Bunke, and Jürgen Schmidhuber. 2009. A novel connectionist system for unconstrained handwriting recognition. Pattern Analysis and Machine Intelligence, IEEE Transactions on. find
  12. Alex Graves, Abdel-rahman Mohamed, and Geoffrey Hinton. 2013a. Speech recognition with deep recurrent neural networks. In Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Conference on. find
  13. Alex Graves, Abdel-rahman Mohamed, and Geoffrey E. Hinton. 2013b. Speech recognition with deep recurrent neural networks. CoRR, abs/1303.5778. — [arXiv:1303.5778](https://arxiv.org/abs/1303.5778) find
  14. Alex Graves, Greg Wayne, and Ivo Danihelka. 2014. Neural turing machines. CoRR, abs/1410.5401. find
  15. Alex Graves. 2013. Generating sequences with recurrent neural networks. CoRR, abs/1308.0850. find
  16. Matthew Henderson, Blaise Thomson, and Steve Young. 2014. Robust dialog state tracking using delexicalised recurrent neural networks and unsupervised adaptation. In Proceedings of IEEE Spoken Language Technology. find
  17. Geoffrey Hinton, Li Deng, Dong Yu, George Dahl, Abdel-rahman Mohamed, Navdeep Jaitly, Andrew Senior, Vincent Vanhoucke, Patrick Nguyen, Tara Sainath, and Brian Kingsbury. 2012. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups. Signal Processi… find
  18. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory. Neural Computation. find
  19. Andrej Karpathy and Li Fei-Fei. 2014. Deep visual-semantic alignments for generating image descriptions. CoRR. find our card
  20. Ravi Kondadadi, Blake Howald, and Frank Schilder. 2013. A statistical nlg framework for aggregated planning and realization. In Proceedings of the 51st Annual Meeting of the ACL. Association for Computational Linguistics. find
  21. Karen Kukich. 1987. Where do phrases come from: Some preliminary experiments in connectionist phrase generation. In Natural Language Generation. Springer Netherlands. find
  22. Irene Langkilde and Kevin Knight. 1998. Generation that exploits corpus-based statistical knowledge. In Proceedings of the 36th Annual Meeting of the ACL, ACL ’98. find our card
  23. François Mairesse and Marilyn A. Walker. 2011. Controlling user perceptions of linguistic style: Trainable generation of personality traits. Computer Linguistics. find
  24. François Mairesse and Steve Young. 2014. Stochastic language generation in dialogue using factored language models. Computer Linguistics. find
  25. François Mairesse, Milica Gašić, Filip Jurčı́ček, Simon Keizer, Blaise Thomson, Kai Yu, and Steve Young. 2010. Phrase-based statistical language generation using graphical models and active learning. In Proceedings of the 48th ACL, ACL ’10. find
  26. William C. Mann and Sandra A. Thompson. 1988. Rhetorical structure theory: Toward a functional theory of text organization. Text. find our card
  27. Tomáš Mikolov and Geoffrey Zweig. 2012. Context dependent recurrent neural network language model. In In Proceedings on IEEE SLT workshop. find
  28. Tomáš Mikolov, Martin Karafit, Lukáš Burget, Jan Černocký, and Sanjeev Khudanpur. 2010. Recurrent neural network based language model. In In Proceedings on InterSpeech. find
  29. Tomáš Mikolov, Stefan Kombrink, Lukáš Burget, Jan H. Černocký, and Sanjeev Khudanpur. 2011a. Extensions of recurrent neural network language model. In ICASSP, 2011 IEEE International Conference on. find
  30. Tomáš Mikolov, Stefan Kombrink, Anoop Deoras, Lukáš Burget, and Jan Černocký. 2011b. Rnnlm - recurrent neural network language modeling toolkit. In In Proceedings on ASRU. find
  31. Danilo Mirkovic and Lawrence Cavedon. 2011. Dialogue management using scripts. EP Patent 1,891,625. find
  32. Alice H. Oh and Alexander I. Rudnicky. 2000. Stochastic language generation for spoken dialogue systems. In Proceedings of the 2000 ANLP/NAACL Workshop on Conversational Systems - Volume 3, ANLP/NAACL-ConvSyst ’00. find
  33. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting on ACL. Association for Computational Linguistics. find our card
  34. Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word representation. In Proceedings of the 2014 Conference on EMNLP. Association for Computational Linguistics. find
  35. Adwait Ratnaparkhi. 2002. Trainable approaches to surface natural language generation and their application to conversational dialog systems. Computer Speech and Language. find
  36. Verena Rieser and Oliver Lemon. 2010. Natural language generation as planning under uncertainty for spoken dialogue systems. In Empirical Methods in Natural Language Generation. Springer-Verlag. find
  37. Mike Schuster and Kuldip K Paliwal. 1997. Bidirectional recurrent neural networks. Signal Processing, IEEE Transactions on. find
  38. Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research. find
  39. Amanda Stent and Martin Molina. 2009. Evaluating automatic extraction of rules for sentence plan construction. In Proceedings of SIGdial. Association for Computational Linguistics. find
  40. Amanda Stent, Rashmi Prasad, and Marilyn Walker. 2004. Trainable sentence planning for complex information presentation in spoken dialog systems. In In Proceedings of the Annual Meeting of the ACL. find
  41. Amanda Stent, Matthew Marge, and Mohit Singhai. 2005. Evaluating evaluation methods for generation in the presence of variation. In in Proceedings of CICLing 2005. find
  42. Martin Sundermeyer, Tamer Alkhouli, Joern Wuebker, and Hermann Ney. 2014. Translation modeling with bidirectional recurrent neural networks. In Proceedings of the 2014 Conference on EMNLP. Association for Computational Linguistics. find
  43. Ilya Sutskever, James Martens, and Geoffrey E. Hinton. 2011. Generating text with recurrent neural networks. In Proceedings of the 28th International Conference on Machine Learning (ICML-11). ACM. find
  44. Ilya Sutskever, Oriol Vinyals, and Quoc V. Le. 2014. Sequence to sequence learning with neural networks. CoRR. find
  45. Marilyn A Walker, Owen C Rambow, and Monica Rogati. 2002. Training a sentence planner for spoken dialogue using boosting. Computer Speech and Language. find
  46. Marilyn Walker, Amanda Stent, Franois Mairesse, and Rashmi Prasad. 2007. Individual and domain adaptation in sentence planning for dialogue. Journal of Artificial Intelligence Research (JAIR. find
  47. Tsung-Hsien Wen, Milica Gašić, Dongho Kim, Nikola Mrkšić, Pei-Hao Su, David Vandyke, and Steve Young. 2015. Stochastic language generation in dialogue using recurrent neural networks with convolutional sentence reranking. In Proceedings of SIGdial. Association for Computational Linguistics. find our card
  48. Paul J Werbos. 1990. Backpropagation through time: what it does and how to do it. Proceedings of the IEEE. find
  49. Kaisheng Yao, Baolin Peng, Yu Zhang, Dong Yu, Geoffrey Zweig, and Yangyang Shi. 2014. Spoken language understanding using long short-term memory neural networks. In In Proceedings on IEEE SLT workshop. IEEE Institute of Electrical and Electronics Engineers. find
  50. Steve Young, Milica Gašić, Blaise Thomson, and Jason D. Williams. 2013. Pomdp-based statistical spoken dialog systems: A review. Proceedings of the IEEE. find
  51. Wojciech Zaremba, Ilya Sutskever, and Oriol Vinyals. 2014. Recurrent neural network regularization. CoRR, abs/1409.2329. find our card
  52. Xingxing Zhang and Mirella Lapata. 2014. Chinese poetry generation with recurrent neural networks. In Proceedings of the 2014 Conference on EMNLP. Association for Computational Linguistics, October. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wen-2015-semantically-conditioned-lstm-based-natural/