An Online Information Retrieval Systems by Means of Artificial Neural Networks

摘要

The aim of this paper is to present a new alternative to the existing Information Retrieval System (IRS) techniques, which are briefly summarized and classified. An IRS prototype has been developed with a technique based on Artificial Neural Networks which are different from those normally used for this type of applications, that is, the self-organising networks (SOM). Two types of network (radial response and multilayer perceptron) are analyzed and tested. It is concluded that, in the case of a limited number of documents and terms, the most suitable solution seems to be the Multilayer Perceptron network. The results obtained with this prototype have been positive, making the possibility of applying this technique in real size cases a cause for a certain degree of optimism.