Intelligent Internet search applications based on VLSI associative processors

Abstract

A general-purpose intelligent search agent on the Internet has been developed using VLSI associative processor (AP) chips as the hardware accelerator. The agent performs similarity-based search in a large volume of database. Although the similarity-based search processing is computationally very expensive, latency-free search has become possible due to the highly parallel maximum-likelihood search architecture of the AP chip. Two search schemes are proposed for intelligent Internet search agents: the server-based solution and the client-based solution. The architecture of the AP chip has been optimized for each of the solutions and their performances are compared. In the server-based solution in which multiple AP chips are mounted on a printed board in the server, 10000 simultaneous accesses to the intelligent search agent are allowed with a response time of less than 1 sec for all visitors. This is by no means possible in software solutions. In the client solution, on the other hand, in which a single AP chip is embedded in a personal computer, a person can enjoy a very comfortable interactive preference search all in a response time of about 10/sup -4/ sec. As an example of similarity-based intelligent search, an E-commerce real-estate agent system has been developed using the AP chip implemented in field-programmable gate arrays (FPGA) and the effectiveness of such agent systems has been demonstrated.