Intelligent workstation agents and unstructured workstation data

Summary form only given. Many researchers envision a future of intelligent, personalized agents in their computer workstations. These agents could provide a variety of personalized assistance helping to manage email, schedule meetings, gather and summarize relevant information, track progress of various projects, etc. Perhaps the greatest barrier to creating such intelligent agents is that the data observable to such an agent is unstructured and difficult to automatically interpret - it includes unstructured workstation files (text, images, and other formats), email, calendar entries, Web accesses, etc. This paper discusses current research toward such intelligent workstation agents, and in particular toward making this unstructured data understandable to computer agents. This research is being conducted as part of a multiuniversity research effort on intelligent personalized assistants.