Survey on Score Normalization : a Case of Result Merging in Distributed Information Retrieval

Merging the outputs of different search engines or information sources in response to a query has been shown to improve performance. In most cases, scores produced by different information sources are not comparable: merging techniques are often segregated into a score normalization step followed by a combination step. The Combination step is usually straight forward and has been an area of active research. However, the normalization step has received less attention; in particular a peculiar attribute such as diversification is largely missing in most Result Merging studies. This survey seeks to explore the various domains of score normalization, especially the results merging phase of a Distributed Information Retrieval environment, and propose a general framework to diversify score normalization through the use of the covariance principle. Key-Words: Machine learning , Classification approach , Distributed Information retrieval , Result merging , Information Retrieval