A probabilistic solution to the selection and fusion problem in distributed information retrieval
Christoph Baumgarten · 1999
A model for optimal information retrieval over a distributed document collection is described and experimentally evaluated. The fusion of retrieval results corresponding to document subcollections is performed according to the Probability Ranking Principle. Part of the model is a selection criterion for effectively limiting the ranking process to a subset of subcollections. 1 Introduction An information retrieval (IR) system is a tool for searching information in a collection of documents that satisfies a user's information need. In order to use an IR system, the user has to formulate his or her information need; the resulting query is then taken by the system to rank the documents according to an estimate of their probabilities of being relevant to the user's information need. For this purpose, a corresponding ranking method is called with the query. It assigns retrieval status values (RSVs) to the documents. A document's RSV determines the position of the document in the ranked lis...