SYSTEM FOR RETRIEVING MULTIMEDIA INFORMATION FROM THE INTERNET USING MULTIPLE EVOLVING INTELLIGENT AGENTS

摘要

A system is provided for retrieving multimedia information using intelligent agents (28) on a computer (10) coupled to a network, such as the Internet, and particularly the World Wide Web (WWW) (15). The system includes a web browser (20) with a graphic user interface (GUI) (21), and an agent server (22) for producing, training and evolving intelligent agents (28). The system includes a natural language processor that determines the subject categories and important terms of a user's query, and of the text of each document retrieved by an agent (28). The agent server embeds an artificial neural network in each agent and trains the neural network in accordance with the natural language query. Each agent (28) processes the subject categories of each document through its neural network and computes a retrieval value for each document. The GUI (21) displays the addresses of the document that have retrieval values above a threshold value. The user or the agent server select which of the retrieved documents are relevant. The agentserver (22) periodically retrains the neural networks to improve their ability to retrieve documents relevant to the query using evolutionary programming techniques. The evolved neural network then replaces the original neural network.