Un Sistema ad Agenti Mobili per la Classificazione e il Recupero Automatico di Documenti

It is presented a system for automatic and concurrent retrieval of documents. The system is based on both the mobile agent paradigm and on a neural network architecture for a sub-symbolic representation of knowledge. The main feature of the system is its versatility in managing documents whose topic class are unknown and not a priori fixed: the system autonomously adapts its document classification capability, exploiting the web directories available in the most common search engines. In this manner it is possible to train a neural classifier for documents retrieval. Experimental results, obtained over 2160 randomly selected documents, confirm the effectiveness of the approach.