A neural based multi-agent system, exploiting the Web Directories as a Knowledge Base for information sharing and documents retrieval, is presented. The system is based on the EαNet architecture, a neural network capable of learning the activation function of its hidden units and having good generalization capabilities. System goal is to retrieve, among documents shared by a networked community, documents satisfying a query and dealing with a specific topic. The system is composed by four agents: the Trainer Agent, the Neural Classifier Agent, the Interface Agent, and the Librarian Agent. The sub-symbolic knowledge of the Neural Classifier Agent is automatically updated each time a new, not included before, document topic is requested by users. The system is very efficient: the experimental results show that, in the best case, a classification error about 10% is obtained.
PILATO, G., SORBELLO, F., CONTI, V., VASSALLO, G., VITABILE, S. (2004). Web Directories as a Knowledge Base to Build a Multi-Agent System for Information Sharing. WEB INTELLIGENCE AND AGENT SYSTEMS, 2, 265-277.
Web Directories as a Knowledge Base to Build a Multi-Agent System for Information Sharing
SORBELLO, Filippo;VASSALLO, Giorgio;VITABILE, Salvatore
2004-01-01
Abstract
A neural based multi-agent system, exploiting the Web Directories as a Knowledge Base for information sharing and documents retrieval, is presented. The system is based on the EαNet architecture, a neural network capable of learning the activation function of its hidden units and having good generalization capabilities. System goal is to retrieve, among documents shared by a networked community, documents satisfying a query and dealing with a specific topic. The system is composed by four agents: the Trainer Agent, the Neural Classifier Agent, the Interface Agent, and the Librarian Agent. The sub-symbolic knowledge of the Neural Classifier Agent is automatically updated each time a new, not included before, document topic is requested by users. The system is very efficient: the experimental results show that, in the best case, a classification error about 10% is obtained.File | Dimensione | Formato | |
---|---|---|---|
16012302.pdf
Solo gestori archvio
Dimensione
159.87 kB
Formato
Adobe PDF
|
159.87 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.