In the last decade, computers have increased the information available to people by many orders of magnitude. The rate of this information explosion is continuing to increase, straining the ability of computer users to comprehend or manage the available information. To help users interact with immense knowledge bases, we propose SemNet, a threedimensional graphical interface. SemNet presents views which allow the users to examine local detail while still maintaining a global representation of the rest of the knowledge base. SemNet also provides many semantic navigation techniques such as relative movement, absolute movement, and teleportation. SemNet is an exploratory research vehicle to address questions such as: How should large data bases of knowledge be presented to users? How should users explore, manipulate, and modify these knowledge bases? Alternative answers to these questions were implemented in SemNet and informally evaluated by using SemNet as an interface to large knowledge bases.