DEPICT: Documents Evaluated as Pictures. Visualizing informationusing context vectors and self-organizing maps

摘要

HNC Software, Inc. has developed a system called DEPICT forvisualizing the information content of large textual corpora. The systemis built around two separate neural network methodologies: contextvectors and self-organizing maps. Context vectors (CVs) are highdimensional information representations that encode the semantic contentof the textual entities they represent. Self-organizing maps (SOMs) arecapable of transforming an input, high dimensional signal space into amuch lower (usually two or three) dimensional output space useful forvisualization. Neither process requires human intervention, nor anexternal knowledge base. Together, these neural network techniques canbe utilized to automatically identify the relevant information themespresent in a corpus, and present those themes to the user in a intuitivevisual form