Giving an Office Assistant Agent a Memory Mechanism.

The extension of ubiquitous computing leads to building networks linking people and services. One problem is to interface humans with such networks. Due to the distributed aspect of the new environment, we advocate to use "intelligent" Personal Assistant agents. Thus, our view of a Personal Assistant (PA) agent is that of an intelligent agent capable of helping users to handle tasks in their workplace. The paper explains how to give a PA agent a memory mechanism, which results in an improved PA agent: MemoPA. Inspired by a case memory model in the domain of Case-Based Reasoning (CBR), we give a PA agent a structured memory. Then, a cognitive memory processor is built to monitor and to update the memory. A first prototype of MemoPA has been developed and implemented on a multi-agent platform named OMAS. The paper indicates how the memory, mechanism makes the PA agent more efficient in handling the tasks requested by the user and in the processing of the user's requests. Finally, several possible improvements for the memory mechanism are discussed