Evaluating Conversational Recommender Systems via User Simulation
This work addresses the time and resource constraints for researchers and developers in conversational information access, though it is incremental as it builds on existing simulation techniques.
The paper tackles the bottleneck of human evaluation in conversational recommender systems by proposing an automated evaluation method using user simulation, showing that preference modeling and task-specific interaction models lead to high correlation with manual assessments.
Conversational information access is an emerging research area. Currently, human evaluation is used for end-to-end system evaluation, which is both very time and resource intensive at scale, and thus becomes a bottleneck of progress. As an alternative, we propose automated evaluation by means of simulating users. Our user simulator aims to generate responses that a real human would give by considering both individual preferences and the general flow of interaction with the system. We evaluate our simulation approach on an item recommendation task by comparing three existing conversational recommender systems. We show that preference modeling and task-specific interaction models both contribute to more realistic simulations, and can help achieve high correlation between automatic evaluation measures and manual human assessments.