GSN: A Graph-Structured Network for Multi-Party Dialogues

Abstract

Existing neural models for dialogue response generation assume that utterances are sequentially organized. However, many real-world dialogues involve multiple interlocutors (i.e., multi-party dialogues), where the assumption does not hold as utterances from different interlocutors can occur ``in parallel.'' This paper generalizes existing sequence-based models to a Graph-Structured neural Network (GSN) for dialogue modeling. The core of GSN is a graph-based encoder that can model the information flow along the graph-structured dialogues (two-party sequential dialogues are a special case). Experimental results show that GSN significantly outperforms existing sequence-based models.

Cite

Text

Hu et al. "GSN: A Graph-Structured Network for Multi-Party Dialogues." International Joint Conference on Artificial Intelligence, 2019. doi:10.24963/IJCAI.2019/696

Markdown

[Hu et al. "GSN: A Graph-Structured Network for Multi-Party Dialogues." International Joint Conference on Artificial Intelligence, 2019.](https://mlanthology.org/ijcai/2019/hu2019ijcai-gsn/) doi:10.24963/IJCAI.2019/696

BibTeX

@inproceedings{hu2019ijcai-gsn,
  title     = {{GSN: A Graph-Structured Network for Multi-Party Dialogues}},
  author    = {Hu, Wenpeng and Chan, Zhangming and Liu, Bing and Zhao, Dongyan and Ma, Jinwen and Yan, Rui},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year      = {2019},
  pages     = {5010-5016},
  doi       = {10.24963/IJCAI.2019/696},
  url       = {https://mlanthology.org/ijcai/2019/hu2019ijcai-gsn/}
}