Mem2Seq: Effectively Incorporating Knowledge Bases into End-to-End Task-Oriented Dialog Systems

TL;DR

Mem2Seq is a novel neural model that effectively integrates knowledge bases into end-to-end task-oriented dialog systems using multi-hop attention and pointer networks, achieving state-of-the-art results.

Contribution

Introduces Mem2Seq, a simple, general, end-to-end differentiable model combining multi-hop attention and pointer networks for knowledge incorporation in dialog systems.

Findings

Mem2Seq trains faster than previous models.

Achieves state-of-the-art performance on three datasets.

Effectively learns correlations between memories.

Abstract

End-to-end task-oriented dialog systems usually suffer from the challenge of incorporating knowledge bases. In this paper, we propose a novel yet simple end-to-end differentiable model called memory-to-sequence (Mem2Seq) to address this issue. Mem2Seq is the first neural generative model that combines the multi-hop attention over memories with the idea of pointer network. We empirically show how Mem2Seq controls each generation step, and how its multi-hop attention mechanism helps in learning correlations between memories. In addition, our model is quite general without complicated task-specific designs. As a result, we show that Mem2Seq can be trained faster and attain the state-of-the-art performance on three different task-oriented dialog datasets.