Understanding the impact of query expansion on federated search

Query expansion (QE) has been studied extensively in traditional search settings due to its efficacy in improving retrieval performance. However, the level of performance achieved in the traditional settings has not been reported in the literature on the federated search. Some of the possible reasons include the lack of complete information regarding the corpus statistics of the databases and their diverse content. Nevertheless, several studies have experimented with different QE approaches and reported mixed results. This paper extends the findings of these publications by studying the impact of using a different source for selecting terms to be used in QE on the federated search. Specifically, the expansion terms are extracted from uniform resource locators (URLs) of the documents returned by each database. The retrieval experiments with TREC 2013 FedWeb dataset demonstrates that the expanded query using the proposed approach performs better in many instances than the unexpanded query.