Abstract
We propose a multi-agent framework integrating effective solutions for document processing. We present an analytical study to ascertain and evaluate the key performance parameters of this integrating multi-agent framework. This is aimed to prove that a framework based on multi-agent technology is indeed an efficient way especially to support solutions integration for solving complex problems, such as document processing, with rather an acceptable response time and speedup.
The diversity of problems, the usefulness of some proposed solutions in the field of document processing and their dependency motivate us to propose this architecture. Our framework agents are cognitive and (or) polyvalent (in terms of abilities). Performances evaluation of the proposed multiagent framework shows that the number of agents to use, the achievable speedup efficiency factors can be easily governed by system parameters such as the task granularity. We found in particular that if the ratio: (Amount of communication and synchronization - Amount of computation) increases 10x then the maximum agents number to use and the maximum speedup factor increase about 3 x without any loss on the efficiency factor.