Metasearch and Federation using Query Difficulty Prediction
Elad Yom‐Tov, Shai Fine, David Carmel, Adam Darlow · 2005
Fusing the results of a distributed Information Retrieval system has often been viewed as two separate problems: metasearch, where several search engines are retrieving from one document collection, and federation, where one search engine is retrieving from several document collections. In this work we present a unified framework for fusion based on predicting query difficulty. We validate our approach using a rigorous, quantitative, experimental procedure, which uses Desktop Search Engines and a standard document collection. We show that the proposed algorithm performs significantly better than previously-proposed algorithms for both metasearch and federation.