Elamur Research⌕
☰
home / papers / 2019 / Interpretable Neural Predictions with Differentiable Binary Variables

Interpretable Neural Predictions with Differentiable Binary Variables

Summary

This ACL 2019 paper addresses how to make neural text classifiers explain themselves. The classifier's decision is based only on a brief, self-sufficient excerpt of the input, called a rationale. The setup follows Lei et al. (2016): one network chooses which words to keep, and a separately parameterised classifier sees only those words. Lei et al. used Bernoulli word masks, which forced them to rely on high-variance REINFORCE training. The authors replace these masks with HardKuma, a new distribution built by stretching a Kumaraswamy variable beyond [0, 1] and clipping it back. Samples can therefore be exactly 0, exactly 1, or in between, and gradients pass through via reparameterisation. Because the distribution has a closed-form CDF, the expected L0 penalty (number of selected words) and a fused-lasso-style transition penalty become differentiable. A Lagrangian relaxation then lets users set the desired fraction of selected text directly. Experiments on beer reviews show results comparable to earlier rationale extractors. HardKuma also works as a sparse attention mechanism for SNLI with the Decomposable Attention model, which matters for transparent, verifiable NLP systems.

Key points

  • HardKuma distribution. A Kumaraswamy(a, b) variable is stretched to an interval (l, r) with l < 0 and r > 1, for example (−0.1, 1.1), then passed through a hard sigmoid. The probability of landing exactly on 0 or 1 is available in closed form.
  • Gradient training. The selector is trained with reparameterised gradients via the inverse CDF of a uniform sample. This avoids REINFORCE and keeps the estimator unbiased, unlike straight-through or SPIGOT.
  • Controlled rationale length. The expected L0 and transition penalties are tractable functions of the selector's parameters. Lagrangian relaxation targets a pre-specified selection rate instead of hand-tuning penalty weights.
  • Tasks and models. Rationale extraction was tested on beer reviews, compared against Lei et al. (2016). Attention experiments used the SNLI corpus with Parikh et al.'s (2016) Decomposable Attention baseline. The provided excerpt omits the exact scores.
  • Code. Released at github.com/bastings/interpretable_predictions.

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

96 papers of our library cite it (of the 140884 whose bibliographies we read); 6 of them have a page here, 90 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/p19-1284

@inproceedings{bastings2019interpretable,
  title = {{Interpretable Neural Predictions with Differentiable Binary Variables}},
  author = {Bastings, Jasmijn and Aziz, Wilker and Titov, Ivan},
  booktitle = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics},
  year = {2019},
  doi = {10.18653/v1/p19-1284},
  url = {https://doi.org/10.18653/v1/p19-1284}
}

References

42 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mark Craven and Jude W Shavlik. 1996. Extracting tree-structured representations of trained networks. In Advances in neural information processing systems, pages 24–30. find
  2. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,… find
  3. Manaal Faruqui, Yulia Tsvetkov, Dani Yogatama, Chris Dyer, and Noah A. Smith. 2015. Sparse overcomplete word vector representations. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (V… find
  4. Leilani H Gilpin, David Bau, Ben Z Yuan, Ayesha Bajwa, Michael Specter, and Lalana Kagal. 2018. Explaining explanations: An overview of interpretability of machine learning. In 2018 IEEE 5th International Conference on Data Science and Advanced Analytics (DSAA), pages 80–89. IEEE. find
  5. David Alvarez-Melis and Tommi Jaakkola. 2017. A causal framework for explaining the predictions of black-box sequence-to-sequence models. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 412– 421. Association for Computational Linguistics. find
  6. Markus Harva and Ata Kaban. 2005. A variational bayesian method for rectified factor analysis. In Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005., volume 1, pages 185–190. IEEE. find
  7. Yoshua Bengio, Nicholas Léonard, and Aaron Courville. 2013. Estimating or propagating gradients through stochastic neurons for conditional computation. arXiv preprint arXiv:1308.3432. find
  8. Aurélie Herbelot and Eva Maria Vecchi. 2015. Building a shared world: mapping distributional to model-theoretic semantic spaces. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 22–32. Association for Computational Linguistics. find
  9. Samuel R. Bowman, Gabor Angeli, Christopher Potts, and Christopher D. Manning. 2015. A large annotated corpus for learning natural language inference. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 632–642. Association for Computational Linguistics. find
  10. Samuel R. Bowman, Jon Gauthier, Abhinav Rastogi, Raghav Gupta, Christopher D. Manning, and Christopher Potts. 2016. A fast unified model for parsing and sentence understanding. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1… find
  11. Qian Chen, Xiaodan Zhu, Zhen-Hua Ling, Si Wei, Hui Jiang, and Diana Inkpen. 2017. Enhanced lstm for natural language inference. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1657–1668. Association for Computational Linguisti… find
  12. Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long Short-Term Memory. Neural Computation, 9(8):1735–1780. find
  13. Jeremy Howard and Sebastian Ruder. 2018. Universal language model fine-tuning for text classification. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 328–339. Association for Computational Linguistics. find
  14. Eric Jang, Shixiang Gu, and Ben Poole. 2017. Categorical reparameterization with gumbel-softmax. International Conference on Learning Representations. find our card
  15. MichaelI. Jordan, Zoubin Ghahramani, TommiS. Jaakkola, and LawrenceK. Saul. 1999. An introduction to variational methods for graphical models. Machine Learning, 37(2):183–233. find
  16. Diederik P. Kingma and Max Welling. 2014. Autoencoding variational bayes. In International Conference on Learning Representations. find
  17. Ponnambalam Kumaraswamy. 1980. A generalized probability density function for double-bounded random processes. Journal of Hydrology, 46(1-2):79–88. find
  18. Tao Lei, Regina Barzilay, and Tommi Jaakkola. 2016. Rationalizing neural predictions. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 107–117. Association for Computational Linguistics. find our card
  19. Zachary Chase Lipton. 2016. The mythos of model interpretability. ICML Workshop on Human Interpretability in Machine Learning (WHI 2016). find
  20. Christos Louizos, Max Welling, and Diederik P Kingma. 2017. Learning sparse neural networks through l 0 regularization. arXiv preprint arXiv:1712.01312. — [arXiv:1712.01312](https://arxiv.org/abs/1712.01312) find
  21. Chris J. Maddison, Andriy Mnih, and Yee Whye Teh. 2017. The concrete distribution: A continous relaxation of discrete random variables. International Conference on Learning Representations. find
  22. Andre Martins and Ramon Astudillo. 2016. From softmax to sparsemax: A sparse model of attention and multi-label classification. In International Conference on Machine Learning, pages 1614–1623. find
  23. Julian McAuley, Jure Leskovec, and Dan Jurafsky. 2012. Learning attitudes and attributes from multi-aspect reviews. In Data Mining (ICDM), 2012 IEEE 12th International Conference on, pages 1020– 1025. IEEE. find
  24. Eric Nalisnick and Padhraic Smyth. 2016. Stick-breaking variational autoencoders. arXiv preprint arXiv:1605.06197. — [arXiv:1605.06197](https://arxiv.org/abs/1605.06197) find
  25. Vlad Niculae and Mathieu Blondel. 2017. A regularized framework for sparse and structured neural attention. In Advances in Neural Information Processing Systems, pages 3338–3348. find
  26. Vlad Niculae, André F. T. Martins, and Claire Cardie. 2018. Towards dynamic computation graphs via sparse latent structure. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 905–911. Association for Computational Linguistics. find
  27. Ankur Parikh, Oscar Täckström, Dipanjan Das, and Jakob Uszkoreit. 2016. A decomposable attention model for natural language inference. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2249–2255. Association for Computational Linguistics. find our card
  28. Hao Peng, Sam Thomson, and Noah A. Smith. 2018. Backpropagating through structured argmax using a spigot. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1863–1873. Association for Computational Linguistics. find
  29. Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. Deep contextualized word representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technolog… find
  30. Danilo Jimenez Rezende, Shakir Mohamed, and Daan Wierstra. 2014. Stochastic backpropagation and approximate inference in deep generative models. In Proceedings of the 31st International Conference on Machine Learning, volume 32 of Proceedings of Machine Learning Research, pages 1278–1286, Bejing, Ch… find our card
  31. Marco Ribeiro, Sameer Singh, and Carlos Guestrin. 2016. “why should i trust you?”: Explaining the predictions of any classifier. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, pages 97–101. Association for Computa… find our card
  32. Jason Tyler Rolfe. 2017. Discrete variational autoencoders. In ICLR. find
  33. Nicholas D. Socci, Daniel D. Lee, and H. Sebastian Seung. 1998. The rectified gaussian distribution. In M. I. Jordan, M. J. Kearns, and S. A. Solla, editors, Advances in Neural Information Processing Systems 10, pages 350–356. MIT Press. find
  34. Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D. Manning, Andrew Ng, and Christopher Potts. 2013. Recursive deep models for semantic compositionality over a sentiment treebank. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1631–… find
  35. Sebastian Thrun. 1995. Extracting rules from artificial neural networks with distributed representations. In Advances in neural information processing systems, pages 505–512. find
  36. Ivan Titov and Ryan McDonald. 2008. A joint model of text and aspect ratings for sentiment summarization. In Proceedings of ACL. find our card
  37. Ronald J Williams. 1992. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine learning, 8(3-4):229–256. find
  38. John Winn and Christopher M Bishop. 2005. Variational message passing. Journal of Machine Learning Research, 6(Apr):661–694. find
  39. Omar Zaidan and Jason Eisner. 2008. Modeling annotators: A generative approach to learning from annotator rationales. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, pages 31–40, Honolulu, Hawaii. Association for Computational Linguistics. find
  40. Omar Zaidan, Jason Eisner, and Christine Piatko. 2007. Using “annotator rationales” to improve machine learning for text categorization. In Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceedings of the Main Confere… find our card
  41. Ye Zhang, Iain Marshall, and Byron C. Wallace. 2016. Rationale-augmented convolutional neural networks for text classification. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 795–804, Austin, Texas. Association for Computational Linguistics. where r… find
  42. Z t 0 −l (22) 1 f K (k; a, b) (r − l)dk (r − l) −∞   0 −l ; a, b . = F K t r−l = Thus we can obtain samples from this generalised-support Kumaraswamy by sampling from a uniform distribution U(0, 1), applying the inverse transform (19), then shifting and scaling the sample according to (20, left).… find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/bastings-2019-interpretable-neural-predictions-differentiable-binary/