Computer Science Department
School of Computer Science, Carnegie Mellon University
Learning Others' Calendars
Senior Honors Thesis
This work develops a method for aiding the process of meeting scheduling through learning about the meetings in he calendars of other users. We assume users do not share their entire calendars. This makes it difficult to determine the exact state of another user's calendar and represent it using a traditional calendar. We solve this problem by representing another agent's calendar as a probability distribution of possible meeting types and present an algorithm called LOC (Learning Others' Calendars)for learning these distributions based on responses to meeting requests. We then present a modification to LOC which uses this information to guide the process of selecting time slots to decrease the number of messages sent during the meeting negotiation process. We implemented these algorithms and ran experiments to test them. We found they successfully learned others' calendars and the second version sent fewer messages than a system which did not leverage the learning information. This shows that calendar learning can aid the scheduling process. Our work integrates into the CMRadar project.