With the WEBSOM method a textual document collection may be organized onto
a graphical map display that provides an overview of the collection and fac
ilitates interactive browsing. Interesting documents can be located on the
map using a content-directed search. Each document is encoded as a histogra
m of word categories which are formed by the self-organizing map (SOM) algo
rithm based on the similarities in the contexts of the words. The encoded d
ocuments an organized on another self-organizing map, a document map, on wh
ich nearby locations contain similar documents. Special consideration is gi
ven to the computation of very large document maps which is possible with g
eneral-purpose computers if the dimensionality of the word category histogr
ams is first reduced with a random mapping method and if computationally ef
ficient algorithms are used in computing the SOMs. (C) 1998 Elsevier Scienc
e B.V. All rights reserved.