Hypertext paradigm in the field of information retrieval: a neural approach

Application of the hypertext paradigm to information retrieval requires 1) an automatic generation of hypertext links, 2) a compact graphical representation of the data. After a brief review of the family of neural algorithms required for deriving a compact and relevant representation of a documentary database, as well as links between synthetic “topics” and documents, we present a user interfaee based on these grounds. This representation is two-step : 1) a global topics map, 2) local topic axes, ranking both terms and documents according to the values of their “centrality index”. A prototype, running in a Macintosh environment and implementing a basic version of this browser, is then described and commented.