Neural Variational Inference for Text Processing

TL;DR

This paper introduces a neural variational inference framework for text processing, enabling improved generative document modeling and question answering by learning stochastic text representations.

Contribution

It presents a novel neural variational inference approach that constructs inference networks conditioned on text, enhancing generative and supervised text models.

Findings

Achieved lowest perplexities on standard document corpora.

Exceeded previous benchmarks in question answering tasks.

Validated framework on diverse text applications.

Abstract

Recent advances in neural variational inference have spawned a renaissance in deep latent variable models. In this paper we introduce a generic variational inference framework for generative and conditional models of text. While traditional variational methods derive an analytic approximation for the intractable distributions over latent variables, here we construct an inference network conditioned on the discrete text input to provide the variational distribution. We validate this framework on two very different text modelling applications, generative document modelling and supervised question answering. Our neural variational document model combines a continuous stochastic document representation with a bag-of-words generative model and achieves the lowest reported perplexities on two standard test corpora. The neural answer selection model employs a stochastic representation layer within an attention mechanism to extract the semantics between a question and answer pair. On two question answering benchmarks this model exceeds all previous published benchmarks.