Self-organization of distributed document archives

Document archives may be regarded as a perfect application arena for unsupervised neural networks because many of the operations computers have to perform on text documents are classification tasks based on noisy patterns. The “noise” originates from the known inaccuracy of mapping free-form natural language to an indexing vocabulary representing the contents of the documents. We describe an approach towards management of distributed document archives based on a system of independent self-organizing maps, each of which represents just a portion of the complete document archive. These individual maps may be integrated in a hierarchical fashion. This leads to enormous flexibility for the user who may define her own personal library that reflects her particular interests