Predicting Retrieval Quality for Resource Selection in Distributed Information Retrieval

In a federated digital library system, it is too expensive to query every accessible library. Resource selection is the task to decide to which libraries a query should be passed. Most existing resource selection algorithms compute a library ranking in a heuristic way. In this paper, we follow a different approach on a better theoretic foundation. The decision-theoretic approach tries to minimise the overall costs of the distributed retrieval. Costs here include—beside retrieval quality—time and money. We present different methods for estimating the retrieval quality of a library. We explore the relationship between the probability of inference and the probability of relevance, and approximate indexing weights with a normal distribution. We also evaluate the different methods on a large test-bed.