Tracking Latent Domain Knowledge - Mapping Scientific Frontiers: The Quest for Knowledge Visualization - page 231

Fig. 7.4 A schematic flowchart of Swanson's Procedure II (Figure 4 reprinted from Swanson and
Smalheiser (1997), available at http://kiwi.uchicago.edu/webwork/fig4.xbm )
In our earlier work, we developed a four-step procedure for visualizing main-
stream domain knowledge (Chen and Paul 2001 ). In particular, the procedure
includes the following four steps:
1. Select highly relevant and highly cited documents from a citation database;
2. Derive citation networks based on the selected population of documents and
simplify citation networks using Pathfinder network scaling;
3. Partition the resultant Pathfinder network according to specialties identified
through Principal Component Analysis;
4. Superimpose the citation history of a document or author over the citation
network.
Our solution to visualizing latent domain knowledge is built upon this four-step
procedure. Instead of simply applying the procedure on highly relevant and highly
cited documents, we incorporate this procedure into a recursive process particularly
suitable for detecting patterns in highly relevant but sparsely cited documents.
Figure 7.6 illustrates the overall strategy of our approach. This approach has three