---
title: "Progressive Open-Domain Response Generation with Multiple Controllable Attributes"
authors:
  - "Haiqin Yang"
  - "Xiaoyuan Yao"
  - "Yiqun Duan"
  - "Jianping Shen"
  - "Jie Zhong"
  - "Kun Zhang"
doi: "10.24963/ijcai.2021/451"
url: "https://ijcai.org/proceedings/2021/451"
pdf: "https://ijcai.org/proceedings/2021/0451.pdf"
venue: "Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence"
year: 2021
pages: "3279-3285"
track: "Main Track"
keywords:
  - "Machine Learning: Learning Generative Models"
  - "Natural Language Processing: Dialogue"
---

# Progressive Open-Domain Response Generation with Multiple Controllable Attributes

Haiqin Yang, Xiaoyuan Yao, Yiqun Duan, Jianping Shen, Jie Zhong, Kun Zhang

*Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence*, pages 3279-3285. DOI: [10.24963/ijcai.2021/451](https://doi.org/10.24963/ijcai.2021/451)

## Abstract

It is desirable to include more controllable attributes to enhance the diversity of generated responses in open-domain dialogue systems.  However, existing methods can generate responses with only one controllable attribute or lack a flexible way to generate them with multiple controllable attributes.  In this paper, we propose a Progressively trained Hierarchical Encoder-Decoder (PHED) to tackle this task.  More specifically, PHED deploys Conditional Variational AutoEncoder (CVAE) on Transformer to include one aspect of attributes at one stage.  A vital characteristic of the CVAE is to separate the latent variables at each stage into two types: a global variable capturing the common semantic features and a specific variable absorbing the attribute information at that stage.  PHED then couples the CVAE latent variables with the Transformer encoder and is trained by minimizing a newly derived ELBO and controlled losses to produce the next stage's input and produce responses as required.  Finally, we conduct extensive evaluations to show that PHED significantly outperforms the state-of-the-art neural generation models and produces more diverse responses as expected.