AISep 22, 2022

Evaluating Agent Interactions Through Episodic Knowledge Graphs

arXiv:2209.11746v2581 citationsh-index: 7
Originality Synthesis-oriented
AI Analysis

This work addresses the challenge of evaluating conversational agents for researchers and developers, but it is incremental as it builds on existing graph-based and evaluation techniques.

The authors tackled the problem of evaluating multimodal conversational agents in open domains by introducing episodic Knowledge Graphs (eKGs) to capture knowledge accumulation over time, resulting in a method that provides more qualitative insights into agent interactions compared to existing metrics.

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.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes