A Sensitivity Analysis of (and Practitioners' Guide to) Convolutional Neural Networks for Sentence Classification

Convolutional Neural Networks (CNNs) have recently achieved remarkably strong performance on sentence classification tasks (Kim, 2014; Kalchbrenner et al., 2014). However, these models require practitioners to specify the exact model architecture and accompanying hyperparameters, e.g., the choice of filter region size, regularization parameters, and so on. It is currently unknown how sensitive model performance is to changes in these configurations for the task of sentence classification. We thus conduct a sensitivity analysis of one-layer CNNs to explore the effect of architecture components on model performance; our aim is to distinguish between important and comparatively inconsequential design decisions for sentence classification. We focus on one-layer CNNs (to the exclusion of more complex models) due to their comparative simplicity and strong empirical performance (Kim, 2014). We derive practical advice from our extensive empirical results for those interested in getting the most out of CNNs for sentence classification. One important observation borne out by our experimental results is that researchers should report performance variances, as these can be substantial due to stochastic initialization and inference.