next up previous
Next: Summary and Future Up: Performance on Hypertext Previous: Computational Time

4.3. Browsing Clusters

  The reordering of rectangular hypertext matrices can be extremely useful in the development of visual browsers for finding related information. Such tools can aid users in locating documents relevant to specific queries in an immediate fashion (i.e., by clusters of hypertext). From Figure 4, for example, we can extract the cluster of articles (see Figure 6) from the letter A of Condensed Columbia Encyclopedia related to people and regions of Persia around 300 BC. Notice that in graph depicted in Figure 6 there are several related articles (shown in blue) not in the collection (i.e., 850 letter A articles) which are links contained in different but related letter A articles: Demosthenes, Diadochi, Greece, Macedon, Peloponnesus, Persia, and Phillip II. This cluster of related hypertext information is fully contained within a subwindow of each of reorderings for the 1778 links by 850 articles CCE-A matrix shown in Figure 7. The display of graphs such as that in Figure 6 coupled with windowing capabilities (e.g., mouse dragging) in a visualization tool for hypertext browsing would be highly effective for scoping the context of large and possibly distributed databases.

 
Figure 6: Graph of Persia-related articles from CCE-A for browsing. 

 
Figure 7: Partitioned windows containing graph of Persia-related CCE-A articles. 

If the entire collection of articles (letters A through Z) of the Condensed Columbia Encyclopedia were distributed across a network (local or even the World-Wide-Web), the graph in Figure 6 as traced by the windows in Figure 7 would allow a user to selectively retrieve foreign or remote documents (e.g., articles from letters B through Z) linked to relevant local documents (e.g., articles from letter A). The relationship of remote documents with both local and other remote documents would be immediately determined by providing a road map of related information across the network. Without such hypertext clustering, related local documents such as Achaea and Arcadia from Figure 6 might be difficult to associate without knowing their common linkage to remote documents such as Peloponnesus and Greece a priori. That is, there would be no need to retrieve the actual texts of Achaea and Arcadia (or their links) to discover their similarity.



next up previous
Next: Summary and Future Up: Performance on Hypertext Previous: Computational Time



Michael W. Berry (berry@cs.utk.edu)
Mon Jan 29 14:30:24 EST 1996