Interpretable Neural Predictions with Differentiable Binary Variables

Interpretable Neural Predictions with Differentiable Binary Variables Jasmijn Bastings ILLC University of Amsterdam bastings@uva.nl Wilker Aziz ILLC University of Amsterdam w.aziz@uva.nl Ivan Titov ILLC, University of Amsterdam ILCC, University of Edinburgh ititov@inf.ed.ac.uk Abstract The success of neural networks comes hand in hand with a desire for more interpretabil- ity. We focus on text classifiers and make them more interpretable by having them provide a justification—a rationale—for their predic- tions. We approach this problem by jointly training two neural network models: a latent model that selects a rationale (i.e. a short and informative part of the input text), and a classifier that learns from the words in the rationale alone. Previous work proposed to assign binary latent masks to input positions and to promote short selections via sparsity- inducing penalties such as L0 regularisation. We propose a latent model that mixes discrete and continuous behaviour allowing at the same time for binary selections and gradient-based training without REINFORCE. In our formu- lation, we can tractably compute the expected value of penalties such as L0, which allows us to directly optimise the model towards a pre- specified text selection rate. We show that our approach is competitive with previous work on rationale extraction, and explore further uses in attention mechanisms. 1 Introduction Neural networks are bringing incredible perfor- mance gains on text classification tasks (Howard and Ruder, 2018; Peters et al., 2018; Devlin et al., 2019). However, this power comes hand in hand with a desire for more interpretability, even though its definition may differ (Lipton, 2016). While it is useful to obtain high classification accuracy, with more data available than ever before it also becomes increasingly important to justify predictions. Imagine having to classify a large collection of documents, while verifying that the classifications make sense. It would be extremely time-consuming to read each document to evaluate the results. Moreover, if we do not pours a dark amber color with decent head that does not recede much . it ’s a tad too dark to see the carbonation , but fairs well . smells of roasted malts and mouthfeel is quite strong in the sense that you can get a good taste of it before you even swallow . Rationale Extractor pours a dark amber color with decent head that does not recede much . it ’s a tad too dark to see the carbonation , but fairs well . smells of roasted malts and mouthfeel is quite strong in the sense that you can get a good taste of it before you even swallow . Classifier look: ⋆⋆⋆⋆ Figure 1: Rationale extraction for a beer review. know why a prediction was made, we do not know if we can trust it. What if the model could provide us the most important parts of the document, as a justification for its prediction? That is exactly the focus of this paper. We use a setting that was pioneered by Lei et al. (2016). A rationale is defined to be a short yet …

Content truncated to 500 words