Variational Neural Discourse Relation Recognizer

Implicit discourse relation recognition is a crucial component for automatic discourse-level analysis and nature language understanding. Previous studies exploit discriminative models that are built on either powerful manual features or deep discourse representations. In this paper, instead, we explore generative models and propose a variational neural discourse relation recognizer. We refer to this model as VIRILE. VIRILE establishes a directed probabilistic model with a latent continuous variable that generates both a discourse and the relation between the two arguments of the discourse. In order to perform efficient inference and learning, we introduce a neural discourse relation model to approximate the posterior of the latent variable, and employ this approximated posterior to optimize a reparameterized variational lower bound. This allows VIRILE to be trained with standard stochastic gradient methods. Experiments on the benchmark data set show that VIRILE can achieve competitive results against state-of-the-art baselines.