On the Evaluation of Dialogue Systems with Next Utterance Classification
This work addresses the problem of evaluating dialogue systems for researchers, providing a validated surrogate task, though it is incremental as it builds on existing NUC proposals.
The paper investigated the relevance of Next-Utterance-Classification (NUC) as an evaluation method for dialogue systems by comparing human and automated performance, finding that humans perform better than chance, with experts outperforming novices and automated systems matching novice levels.
An open challenge in constructing dialogue systems is developing methods for automatically learning dialogue strategies from large amounts of unlabelled data. Recent work has proposed Next-Utterance-Classification (NUC) as a surrogate task for building dialogue systems from text data. In this paper we investigate the performance of humans on this task to validate the relevance of NUC as a method of evaluation. Our results show three main findings: (1) humans are able to correctly classify responses at a rate much better than chance, thus confirming that the task is feasible, (2) human performance levels vary across task domains (we consider 3 datasets) and expertise levels (novice vs experts), thus showing that a range of performance is possible on this type of task, (3) automated dialogue systems built using state-of-the-art machine learning methods have similar performance to the human novices, but worse than the experts, thus confirming the utility of this class of tasks for driving further research in automated dialogue systems.