Favorite+: Favorite Tuples Extraction via Regret Minimization

When faced with a database containing millions of tuples, a user might be only interested in some of them. In this paper, we study how to help an end user to find the favorite tuples based on the recent advancements in regret minimization queries, which guarantees the tuples returned are not far from the user's favorite tuple in the database, without asking the user to scan the entire database. We consider three types of regret minimization queries: (1) End-to-end query: Given an output size k, we directly return a subset of at most k tuples from the database; (2) Interactive query: We identify the user's favorite tuple via user interaction, where a user might be presented with a few pairs of tuples, and the user is asked to indicate the one s/he favors more from each pair; and (3) Incremental query: Analogous to how we use search engines, if the user is not satisfied with the current tuples, we continually return more. We developed a demonstration system, called Favorite+, by supporting the above queries. We demonstrate that the system could help the users to find their favorite tuples in the database efficiently and effectively.