Design and implementation of ASSIST : Automatic Secretary System using Internet and Smart-agent Technologies

Abstract

With the advent of various Internet technologies, it is now feasible and cost-effective to provide secretarial services by electronic means in general and by software agents in particular. This thesis presents an automatic secretary system called ASSIST (Automatic Secretary System using Internet and Smart-agent Technologies). Generally, ASSIST provides basic secretarial services such as meeting scheduling, email management and document management for professionals and executives. Like other electronic secretarial systems, our focus is on the scheduling of meetings. Inspired by other similar systems, the meeting scheduling mechanism basically works as follows. Upon receiving a request, some possible time slots are found based on information about the meeting. The system then sends these time slots to the invitees enquire as to their availability. The meeting can be arranged if a mutually agreed time slot is found. If not, the above process is repeated. In most of the existing systems, an arbitrary number of time slots are used for scheduling purposes. In this thesis, we present two analytical models for calculating the number of suggested time slots so that the average cost of scheduling a meeting can be minimized. Based on the first model, the best number of time slots per scheduling cycle can be determined. As an extension to the first model, the second model is Markov-decision-based, which takes into account the scheduling deadline and other meeting parameters. By using this model, the number of suggested time slots can be varied dynamically. For both models, we have conducted extensive analysis to evaluate their performance. Finally, we have investigated some matching algorithms for finding a common time slot among some proposed time slots. These algorithms are the random matching algorithm, ascending availability matching algorithm, descending availability matching algorithm, simplified ascending availability matching algorithm and simplified descending availability matching algorithm. We have compared the aforementioned algorithms by using computer simulations. It was found that the ascending availability matching algorithm provides the best performance in general in terms of the scheduling cost and the successful rate for finding a time slot.