Neural approvals to conversational AI (1)

Question Answering, Task-oriented Dialogues and Social Chatbots

This week we will read Jianfeng Gao、Michel Galley and Lihong Li A paper written jointly

---

Abstract

1. Overview of conversations in recent years AI Neural methods for domain development
2. Divide the dialogue system into three categories , A review of the best neural methods for each class with specific systems and models , Discussion is different from traditional methods , The achievements we have made and the challenges we still need to face

---

brief introduction

The author comes straight to the point , Intelligent dialogue system should not only imitate human dialogue, but also answer various types of questions , This is the long-term goal of intelligent dialogue system , Fortunately, in-depth learning 、 Strengthen learning and apply it to conversation AI, Adding a large amount of data to the training process makes it promising to achieve this goal in academia .

This paper is an extension and supplement to the previous work , Not only in data and supervised learning , And this is also the first overview in conversation AI A neural approach to , On the target object NLP and IR The researchers .

The contribution of this paper lies in ：

1. The overview includes a unified view from the optimization decision