Authoritative Re-Ranking in Fusing Authorship-Based Subcollection Search Results

We examine the use of authorship information to divide IR test collections into subcollections and apply techniques from the field of distributed information retrieval to enhance the baseline search results. We determine the expertise of each author, based on the content of their documents, and use this knowledge to construct rankings of the different author subcollections for each query. We go on to demonstrate that these rankings can then be used to re-rank baseline search results and improve performance significantly. We also perform experiments in which we base expertise ratings only on first authors or on all except the final authors and find that these limitations do not further improve our re-ranking method.