Rapid Knowledge Formation

The goal of the DARPA-funded Rapid Knowledge Formation (RKF) project (1999-2003) was to develop technology that enables people who are untrained in AI to build knowledge bases efficiently and accurately. This goal aligned perfectly with our research agenda, which we're continuing to explore under Project Halo.

As part of the SRI team of researchers and engineers working on the RKF project, we developed the SHAKEN system for capturing domain knowledge from Subject Matter Experts (SMEs).

A SME builds a knowledge base using SHAKEN by creating and editing concept maps. (Here's a good introduction to concept maps.) The nodes and relations in the map are drawn from our Component Library, a comprehensive collection of general events and entities. We claim that relatively few components, perhaps a few thousand, are sufficient for SMEs to assemble models of virtually any physical mechanism or process. We claim that these components are independent of domain, and that assembly from components instantiated to a domain is a natural way for SMEs to create knowledge-base content.


An independent group (IET) evaluated SHAKEN, the Component Library and its associated claims. IET hired a group of four biologists who were given a short course on using the technology, then asked to independently build a Knowledge System covering the content of one section of a college-level textbook on cell biology. The Knowledge Systems were evaluated by posing a set of questions, and the answers were judged by another biologist to be "mostly correct" (see the first publication, below, for details). These results are significant: they suggest that the basic machinery works, providing a method for knowledge acquisition without the users having to encode axioms directly (or even encounter terms like "concept", "relation", "instance", "quantification", etc), and that the axioms assembled from pre-built components support automated reasoning and question answering.

Contact SRI to use SHAKEN or read documentation on it.

Key Publications

Created by Bruce Porter
Maintained by Dan Tecuci
Last modified April 19, 2006