The synthesis of specialty narratives from co-citation clusters

A method is described for generating reviews or synopses of scientific fields called specialty narratives. The raw data for the narrative are the Science Citation Index and selected texts from the published scientific literature. The review process is modeled as a walk through a co-citation network using a quasi-minimal spanning tree path and a depth-first search method. Text for the narrative is derived from a statistical analysis of passages which cite the core (highly cited) documents in the cluster. A method is devised for selecting the passage which is most representative of the citing passages for a specific core document using shared content words. This is called the consensus passage. Transitions between nodes in the network are derived from an analysis of co-citing contexts, which reveal why the concepts represented by the documents are associated. The final specialty narrative consists of consensus passages ordered by the tree search, and connected to one another by transitional sentences. The method is applied to a co-citation cluster in the field of cancer virology, and the resulting specialty narrative is described in relation to the structure of the original co-citation map. Implications of the procedure for the cognitive processes involved in reviewing a field, and the extension of the method to other fields and higher level maps of science are discussed.