arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1804.08217cs.CL
收藏

Mem2Seq:有效地将知识库转换为面向任务的端到端对话系统

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

Human Language Technology Center(人类语言技术中心);Center for Artificial Intelligence Research (CAiRE)(人工智能研究中心(CAiRE));Department of Electronic and Computer Engineering(电子与计算机工程系);The Hong Kong University of Science and Technology(香港科技大学)

Andrea Madotto, Chien-Sheng Wu, Pascale Fung

摘要:

端到端的面向任务的对话系统通常遭受的挑战,纳入知识库。在本文中,我们提出了一种新颖而简单的端到端可微模型,称为记忆到序列(Mem2Seq)来解决这个问题。Mem2Seq是第一个将记忆上的多跳注意力与指针网络思想相结合的神经生成模型。我们经验性地展示了Mem2Seq如何控制每个生成步骤,以及它的多跳注意力机制如何帮助学习记忆之间的相关性。此外,我们的模型是相当普遍的,没有复杂的特定任务的设计。结果表明,Mem2Seq可以更快地训练,并在三个不同的面向任务的对话数据集上达到最先进的性能。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑