The boomerang effect: retrieving scientific documents via the network of references and citations

The theory of polyrepresentation [2, 3] provides a theoretical background for how to exploit such features. In summary, the theory hypothesises that overlaps between different cognitive representations of both users’ information needs as well as documents can be exploited for reducing the uncertainties inherent in IR, and thereby improve the performance of IR systems. Two or more different cognitive representations pointing at the same documents is regarded as multi-evidence of those documents being relevant, and suggests to apply a principle of ‘intentional redundancy’ [2] with the purpose of reducing the uncertainties by placing emphasis on overlaps between representations. Better results are expected when cognitively unlike representations are used, e.g., the document title (made by the author) vs. intellectually assigned descriptors from indexers.