Information retrieval in distributed hypertexts

Hypertext is a generalization of the conventional linear text into a non-linear text formed by adding cross-reference and structural links between different pieces of text. A hypertext can be regarded as an extension of a textual database by adding a link structure among the different text objects it stores. We present a tool for finding information in a distributed hypertext such as the World-Wide Web (WWW). Such a hypertext is a distributed textual database in which text objects residing at (the same and) different sites have links to each other. In such a database retrieval is limited to the transfer of documents with a known name. Names of documents serve as links between different documents, and finding such references names is only possible by parsing documents that have embedded links to other documents.

Full-text search in such hypertexts is not feasible because of the discrepancy between the large size of the hypertext and the relatively low bandwidth of the network. We present an information retrieval algorithm for distributed hypertexts, which does an incomplete search through a part of the hypertext. Heuristics determine the selection of the documents that are to be retrieved and searched. A prototype implementation for the WWW, on top of Mosaic for X, is being used by an increasingly large user base.