Evaluating agent interactions through episodic knowledge graphs

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

We present a new method based on episodic Knowledge Graphs (eKGs) for evaluating (multimodal) conversational agents in open domains. This graph is generated by interpreting raw signals during conversation and is able to capture the accumulation of knowledge over time. We apply structural and semantic analysis of the resulting graphs and translate the properties into qualitative measures. We compare these measures with existing automatic and manual evaluation metrics commonly used for conversational agents. Our results show that our Knowledge-Graph-based evaluation provides more qualitative insights into interaction and the agent’s behavior.
Original languageEnglish
Title of host publicationProceedings of the 1st Workshop on Customized Chat Grounding Persona and Knowledge @ COLING2022
PublisherAssociation for Computational Linguistics (ACL)
Pages15-28
Number of pages14
DOIs
Publication statusPublished - 2022

Publication series

NameCOLING
PublisherACL
Number13
Volume29

Bibliographical note

BEST PAPER AWARD

Fingerprint

Dive into the research topics of 'Evaluating agent interactions through episodic knowledge graphs'. Together they form a unique fingerprint.

Cite this