Best Bets: Thousands Queries in Search of a Client
A number of applications require selecting targets for specific contents on the basis of criteria defined by the contents providers themselves. Performing this task requires inverting the direction of search, retrieving queries from documents rather than vice versa, as in ordinary information retrieval. We present a retrieval system called Best Bets that handles this case. Best Bets generalize Information Filtering and encompass a variety of applications including editorial suggestions, promotional campaigns and targeted advertising. In particular we show how to model Google AdWords™ as a Best Bets system. We discuss how to address performance issues in the implementation of Best Bets, in particular efficient search, incremental updates and dynamic ranking. We report about experiments on an implementation of our techniques and a comparison with previous approaches.