Scone

Overview

Scone provides support for representing symbolic knowledge about the world, whether this is general "common sense" knowledge or knowledge about some specific application domain. For Radar, Scone must represent knowledge about many things, including people, rooms, events, equipment, vendors, caterers, and tasks. For each of these types, Scone can represent subtypes: people may be faculty or students, conference speakers or organizers, VIPs, or vegetarians, or combinations of these types. Scone also represents attributes: for a room, we may know its capacity, whether it has a projector, information about its availability and cost, and so on. Scone must be able to handle partial or missing knowledge. It must also tolerate some amount of confusion and inconsistency, since its knowledge will come from many sources.

Like other KB systems, Scone also provides support for simple "common sense" inference at query-time. If I tell Scone "John is an research professor at Carnegie Mellon University", Scone can infer that John is human, is a faculty member, and that he is therefore allowed to reserve certain meeting rooms.

In addition to supporting inference, Scone provides support for search within the knowledge base. For example, it can handle queries like "Are there any full professors working on the Radar project whose home department is HCII?"

Technical Details

Scone differs from other knowledge-base systems in the way it implements this search and inference. Scone uses marker-passings algorithms originally designed for a hypothetical massively parallel machine (the NETL machine). These marker-passing algorithms do not have the full inference power of a general-purpose theorem-prover, but they are very fast and can handle the kinds of search and inference that are needed for most "common sense" reasoning. Scone has a powerful "context" mechanism that can be used to reason efficiently about various states of the world and "what if" scenarios.

The greatest problem for users of current KB systems has been the difficulty of adding new knowledge to the system and making that knowledge fully effective. A major focus of our research on Scone is to make it easy for users with no special training to add new knowledge to the Scone KB. Several graduate students in the Scone group are working on extending the natural-language capabilities of Scone, so that we can communicate with Scone in simple English, perhaps via Email.

Our plan is to release Scone - the Common Lisp code, a relatively small "core" knowledge base, and a programmer-level manual - as open-source software. This release will be followed by periodic updates as we continue to develop the Scone engine and associated knowledge bases. Scott Fahlman is working on a tutorial book that should make it much easier for beginners to make use of the Scone software in projects of their own. We hope that this will lead to an active worldwide community of Scone users who will extend the system in various ways and who will develop open-source knowledge bases for many domains.