Estimating the relevance of a data source using a fuzzy-cardinality-based summary

In this paper, we consider the situation where a fuzzy query is submitted to distributed data sources. In order to save bandwith and processing cost, we propose a technique whose aim is to forward the query to the most relevant sources only. It is assumed that a fuzzy summary of every data source is available, and the approach we propose consists in estimating the relevance of a source wrt to a user query, based on its associated summary. The general case where the user does not necessarily employ the vocabulary (i.e., the labels from the fuzzy partitions) that was used for summarizing the source is considered.