Merging Results From Isolated Search Engines

Two new techniques for merging search results are introduced: Feature Distance ranking algorithms and Reference Statistics. These techniques are compared with other published methods, using TREC eeectiveness evaluations based on human relevance judgements and input rankings from 5 diierent search engines over 5 disjoint document collections. The new techniques are found to be more eeective than existing methods in an isolated-server environment such as the World Wide Web. In addition, Feature Distance algorithms are found to be as eeective in an isolated-server environment using Reference Statistics as they are in an integrated-server environment.