Home | Business News | Browse by Publication | X | XML Journal

Distributed AI, schedules, and the Semantic Web: RCal provides a glimpse of what''s to come. (Automated Scheduling).

Article Excerpt
Meetings are an integral part of modern life, regardless of whether they're formally established at the workplace or casual agreements made over the phone; Many factors affect meeting scheduling, some of which are explicit (What existing meetings do I have?), implicit (I prefer to avoid meetings before 10:00 am), or cultural (social events should be scheduled for Friday evenings and weekends, but not Sunday mornings).

In addition, humans typically reason about where the meeting will take place, its duration, purpose, and so on. All these factors contribute to making meeting scheduling a difficult problem. However, recent developments in automated preference acquisition, multiagent negotiation, and reasoning techniques for semantic content on the Web are slowly making automated scheduling a reality.

An important factor in automated meeting scheduling is being able to share, understand, and reason about all the details of an event or meeting request. The Semantic Web facilitates the representation and distribution of knowledge as structured data with meaning, thus allowing agents to reason about concepts in the real world. We've developed several intelligent agents that negotiate with each other to organize meetings on behalf of their users, using published knowledge to make appropriate decisions. The RETSINA Semantic Web Calendar Agent (RCal) is one such agent built using the RETSINA AI infrastructure, which augments a widely used Personal Information Manager (PIM)--MS Outlook 2000. RCal combines knowledge about its user's current schedule, information about colleagues and friends (using MS Outlook 2000's Contact entries), and knowledge gathered from the Semantic Web to better automate meeting scheduling and management. In this scenario, users running MS Outlook 2000 on their desktops also have an instance of RCal running in the background, acting on their behalf.

RCal (see Figure 1) schedules meetings for its user, updates the user's calendar with schedules from the Semantic Web, interacts with Web services that may provide additional relevant information pertaining to scheduled meetings, and provides alerts based on occurring events.

[FIGURE 1 OMITTED]

Distributed Meeting Scheduling

RCal negotiates with other agents to find mutually agreeable times based on the user's availability and preferences. Traditionally, the burden of maintaining an up-to-date calendar has fallen on the user--a task that is time consuming and error prone. To address this, RCal can reason about events and schedules published on the Semantic Web, and automatically incorporate them directly into the user's schedule. This reduces the burden on the user, and maintains an up-to-date calendar that can be consulted by the agent when scheduling meetings.

RCal currently supports two types of distributed meeting negotiation--multiparty negotiation and appointment-request negotiation. Multiparty negotiation occurs when several agents try to identify a mutually agreeable meeting slot based on their users' current schedules and preferences, whereas appointment-request negotiation identifies possible meeting times for one party based on a meeting request. This latter form of negotiation is used by the Web-based E-Secretary to allow people to request meetings or appointments via a Web-based interface.

RCal's multiparty negotiation occurs when someone desires a meeting with one or more individuals, each of which employ their own RCal agents to manage their schedules. RCal goes through several rounds of automated negotiation with other RCal agents until all the agents agree upon a common time. Users can instruct RCal to schedule a meeting by specifying a particular time they would prefer or...