Implicit Deep Latent Variable Models for Text Generation
This work addresses a key bottleneck in text generation models for NLP researchers, offering an incremental improvement over existing VAEs.
The paper tackled the limited representation power and posterior collapse issues in variational auto-encoders for text generation by proposing implicit deep latent variable models with sample-based variational distributions and a regularization to match aggregated posterior to prior, achieving improved performance in tasks like language modeling, style transfer, and dialog generation.
Deep latent variable models (LVM) such as variational auto-encoder (VAE) have recently played an important role in text generation. One key factor is the exploitation of smooth latent structures to guide the generation. However, the representation power of VAEs is limited due to two reasons: (1) the Gaussian assumption is often made on the variational posteriors; and meanwhile (2) a notorious "posterior collapse" issue occurs. In this paper, we advocate sample-based representations of variational distributions for natural language, leading to implicit latent features, which can provide flexible representation power compared with Gaussian-based posteriors. We further develop an LVM to directly match the aggregated posterior to the prior. It can be viewed as a natural extension of VAEs with a regularization of maximizing mutual information, mitigating the "posterior collapse" issue. We demonstrate the effectiveness and versatility of our models in various text generation scenarios, including language modeling, unaligned style transfer, and dialog response generation. The source code to reproduce our experimental results is available on GitHub.