Interpretable Structure Induction via Sparse Attention

Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, pages 365–367 Brussels, Belgium, November 1, 2018. c⃝2018 Association for Computational Linguistics 365 Interpretable Structure Induction Via Sparse Attention Ben Peters† Vlad Niculae† and Andr´e F.T. Martins†‡ †Instituto de Telecomunicac¸˜oes, Lisbon, Portugal ‡Unbabel, Lisbon, Portugal benzurdopeters@gmail.com, vlad@vene.ro, andre.martins@unbabel.com. 1 Introduction Neural network methods are experiencing wide adoption in NLP, thanks to their empirical per- formance on many tasks. Modern neural ar- chitectures go way beyond simple feedforward and recurrent models: they are complex pipelines that perform soft, differentiable computation in- stead of discrete logic. Inspired by pioneering work by, e.g. Kohonen et al. (1981); Das et al. (1992); Schmidhuber (1992), such modern dif- ferentiable architectures include neural memories (Sukhbaatar et al., 2015) and attention mecha- nisms (Bahdanau et al., 2015). The price of such soft computing is the introduction of dense depen- dencies, which make it hard to disentangle the pat- terns that trigger a prediction. Our recent work on sparse and structured latent computation (Mar- tins and Astudillo, 2016; Niculae and Blondel, 2017; Niculae et al., 2018; Malaviya et al., 2018) presents a promising avenue for enhancing inter- pretability of such neural pipelines. Through this extended abstract, we aim to discuss and explore the potential and impact of our methods. The principle of parsimony suggests that sim- pler explanations are more plausible and inter- pretable. Our perspective is similar to prior work on regularizing model weights (Hastie et al., 2015), but with a twist: instead of model sparsity that tells us which “static” groups of variables are relevant for a task, we now have a “dynamic” form of sparsity that tells us, for a particular input ob- ject, where we should attend to produce a decision. • sparsity: shrinking probabilities to zero to prune entire parts of the input when explaining a prediction (Martins and Astudillo, 2016); • regularization: injecting prior assumptions, such as that neighbouring words should be fused together (Niculae and Blondel, 2017); • constraints: constraining probabilities within lower and upper bounds, to prevent words from receiving too much or too little attention (Malaviya et al., 2018); • structure: learning latent structure predictors (e.g. aligners or parsers), to induce a compact representation as a small, interpretable set of global structures (Niculae et al., 2018). 2 Attention Mechanisms The key background for our work is the concept of attention. Attention mechanisms and mem- ory networks are able to “point” to relevant items (e.g. words or pixels) that determine the final pre- diction, approximating a discrete choice (argmax) with a soft, differentiable one (softmax). Let H = [h1, . . . , hL] ∈ RD×L be a matrix whose columns are vectors encoding the L different choices (for example, words in a sentence). An attention mech- anism maps a H and a control state s to a proba- bility distribution p ∈ △L over the L choices.1 This can be split into (i) generating scores for each choice, e.g., zi = v⊤tanh(Whi + Us) for …

Content truncated to 500 words