Posterior-GAN: Towards Informative and Coherent Response Generation with Posterior Generative Adversarial Network

TL;DR

Posterior-GAN introduces a novel adversarial framework that leverages query-response-future turn triples to generate more informative and coherent responses in neural conversational models, addressing the issues of dull responses and loose coupling.

Contribution

The paper proposes a new encoder-decoder generative adversarial network with dual discriminators to improve response informativeness and coherence by considering future conversation context.

Findings

Enhanced response informativeness demonstrated in experiments

Improved coherence of generated responses verified by human evaluation

Outperforms baseline models on automatic metrics

Abstract

Neural conversational models learn to generate responses by taking into account the dialog history. These models are typically optimized over the query-response pairs with a maximum likelihood estimation objective. However, the query-response tuples are naturally loosely coupled, and there exist multiple responses that can respond to a given query, which leads the conversational model learning burdensome. Besides, the general dull response problem is even worsened when the model is confronted with meaningless response training instances. Intuitively, a high-quality response not only responds to the given query but also links up to the future conversations, in this paper, we leverage the query-response-future turn triples to induce the generated responses that consider both the given context and the future conversations. To facilitate the modeling of these triples, we further propose a novel encoder-decoder based generative adversarial learning framework, Posterior Generative Adversarial Network (Posterior-GAN), which consists of a forward and a backward generative discriminator to cooperatively encourage the generated response to be informative and coherent by two complementary assessment perspectives. Experimental results demonstrate that our method effectively boosts the informativeness and coherence of the generated response on both automatic and human evaluation, which verifies the advantages of considering two assessment perspectives.