We present Cheshire, a recommendation system to help in building reading lists for topic shifts. Given a document collection, a starting topic and a target topic (expressed by keywords), Cheshire recommends the sequence of documents that bridges the gap between input and target topics with the smallest difference in content between each document. To do that, the document collection is represented as a graph, where documents are nodes related by weighted edges. Edges are created whenever a set

more »

... words is shared by two documents. In this paper we present experiments with different methods for choosing words that create edges, and the weights to be assigned to each edge. Results are evaluated by comparison with a dummy baseline and with a manually created gold standard.