Exquisitor at the Video Browser Showdown 2024: Relevance Feedback Meets Conversational Search

An important open problem in video retrieval and exploration concerns the generation and refinement of queries for complex tasks that standard methods are unable to solve, especially when the systems are used by novices. In conversational search, the proposed approach is to ask users to interactively refine the information provided to the search process, until results are satisfactory. In user relevance feedback (URF), the proposed approach is to ask users to interactively judge query results, which in turn refines the query used to retrieve the next set of suggestions. The question that we seek to answer in the long term is how to integrate these two approaches: can the query refinements of conversational search directly impact the URF model, and can URF judgments directly impact the conversational search? We extend the existing Exquisitor URF system with conversational search. In this version of Exquisitor, the two modes of interaction are separate, but the user interface has been completely redesigned to (a) prepare for future integration with conversational search and (b) better support novice users.