Speaker Role Contextual Modeling for Language Understanding and Dialogue Policy Learning

This paper is published in Proceedings of The 8th International Joint Conference on Natural Language Processing (IJCNLP 2017).

Full paper: arXiv, ACL

Code: GitHub

Cite the paper:

@inproceedings{chi2017speaker,
  author    = {Ta-Chung Chi and Po-Chun Chen and Shang-Yu Su and Yun-Nung Chen},
  title	    = {Speaker Role Contextual Modeling for Language Understanding and Dialogue Policy Learning},
  booktitle = {Proceedings of IJCNLP},
  year	    = {2017}
}

Language understanding (LU) and dialogue policy learning are two essential components in conversational systems. Human-human dialogues are not well-controlled and often random and unpredictable due to their own goals and speaking habits.

This paper proposes a role-based contextual model to consider different speaker roles independently based on the various speaking patterns in the multi-turn dialogues. The experiments on the benchmark dataset show that the proposed role-based model successfully learns role-specific behavioral patterns for contextual encoding and then significantly improves language understanding and dialogue policy learning tasks (The source code is available at: https://github.com/MiuLab/Spk-Dialogue).