CLOct 8, 2020

GRADE: Automatic Graph-Enhanced Coherence Metric for Evaluating Open-Domain Dialogue Systems

arXiv:2010.03994v11018 citations
Originality Incremental advance
AI Analysis

This work addresses the challenge of evaluating open-domain dialogue systems, which is crucial for developing high-quality conversational AI, though it is incremental as it builds on existing graph-based and representation methods.

The authors tackled the problem of automatically evaluating dialogue coherence by proposing GRADE, a metric that incorporates both utterance-level and topic-level graph representations, which significantly outperformed existing state-of-the-art metrics in terms of Pearson and Spearman correlations with human judgments.

Automatically evaluating dialogue coherence is a challenging but high-demand ability for developing high-quality open-domain dialogue systems. However, current evaluation metrics consider only surface features or utterance-level semantics, without explicitly considering the fine-grained topic transition dynamics of dialogue flows. Here, we first consider that the graph structure constituted with topics in a dialogue can accurately depict the underlying communication logic, which is a more natural way to produce persuasive metrics. Capitalized on the topic-level dialogue graph, we propose a new evaluation metric GRADE, which stands for Graph-enhanced Representations for Automatic Dialogue Evaluation. Specifically, GRADE incorporates both coarse-grained utterance-level contextualized representations and fine-grained topic-level graph representations to evaluate dialogue coherence. The graph representations are obtained by reasoning over topic-level dialogue graphs enhanced with the evidence from a commonsense graph, including k-hop neighboring representations and hop-attention weights. Experimental results show that our GRADE significantly outperforms other state-of-the-art metrics on measuring diverse dialogue models in terms of the Pearson and Spearman correlations with human judgements. Besides, we release a new large-scale human evaluation benchmark to facilitate future research on automatic metrics.

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