Proceedings of the International Workshop on Applications of Neural Networks to Telecommunications

Multistage Information Filtering Using Cascaded Neural Networks

Thomas John

Southwestern Bell Technology Resources St. Louis, MO. [e-mail]

Abstract

This paper describes a multistage neural network classifier for filtering information. Classifiers based on adaptive mapping networks are used to incrementally refine information as it passes through a cascade. Results are derived showing that such a refinement process can be performed. Experimental results are derived to determine the computational efficiency and accuracy of the filter cascade. Applications to information distribution and information bases are suggested.

1  Introduction

Although there has been much work on creating information retrieval systems and information filters ...