Sharding for literature search via cutting citation graphs

Distributed information retrieval will be a general practice of searching in the exponentially growing scientific literature. At the core of the success for efficient and effective distributed literature search is an adequate sharding policy. This paper proposes a novel sharding policy for literature search that bases on cutting the document citation and co-citation graphs. Experiments on the iSearch test collection reveal that relevant documents for a given query distribute over the shards generated through citation graph cutting in such a pattern that a few shards becomes optimal shards thus can be leveraged in a selective search strategy, potentially leading to efficient and effective literature search solutions.