FineD-Eval: Fine-grained Automatic Dialogue-Level Evaluation
This work addresses the need for more comprehensive dialogue evaluation metrics for researchers and developers in conversational AI, though it is incremental as it builds on existing model-based metrics.
The paper tackles the problem of evaluating open-domain dialogues by proposing a multi-dimensional, dialogue-level metric that assesses multiple quality dimensions, achieving around 16% relative improvement over existing state-of-the-art metrics on benchmarks.
Recent model-based reference-free metrics for open-domain dialogue evaluation exhibit promising correlations with human judgment. However, they either perform turn-level evaluation or look at a single dialogue quality dimension. One would expect a good evaluation metric to assess multiple quality dimensions at the dialogue level. To this end, we are motivated to propose a multi-dimensional dialogue-level metric, which consists of three sub-metrics with each targeting a specific dimension. The sub-metrics are trained with novel self-supervised objectives and exhibit strong correlations with human judgment for their respective dimensions. Moreover, we explore two approaches to combine the sub-metrics: metric ensemble and multitask learning. Both approaches yield a holistic metric that significantly outperforms individual sub-metrics. Compared to the existing state-of-the-art metric, the combined metrics achieve around 16% relative improvement on average across three high-quality dialogue-level evaluation benchmarks.