Towards a Dialogue System with Long-term, Episodic Memory

Daniel Kondratyuk, Casey Kennington


Intelligent personal assistants lack long-term memory. We propose graph databases as a extensible solution to this problem by representing relevant knowledge as entities, properties, and relations between them. We demonstrate through two experiments that our approach lends itself to a system that can improve natural language understanding by updating its knowledge dynamically in a generalizable and interpretable fashion.


Cite as: Kondratyuk, D., Kennington, C. (2017) Towards a Dialogue System with Long-term, Episodic Memory. Proc. SEMDIAL 2017 (SaarDial) Workshop on the Semantics and Pragmatics of Dialogue, 152-153.


@inproceedings{Kondratyuk2017,
  author={Daniel Kondratyuk and Casey Kennington},
  title={Towards a Dialogue System with Long-term, Episodic Memory},
  year=2017,
  booktitle={Proc. SEMDIAL 2017 (SaarDial) Workshop on the Semantics and Pragmatics of Dialogue},
  pages={152--153}
}