Semantic Retrieval and Classification of Web Pages Using Frames and Neural Networks

Abstract

Vectors of Hilbert space are used to represent (hyper) texts. Instead of a basis of the space, a frame is constructed so as to represent particular semantic units that could be relevant for a given specific domain. It is suggested to use Kohonen’s neural networks for unsupervised classification of texts. A special emphasis is put on handling information saved in the WWW.