The Stony Brook SYNCHEM system is a large knowledge-based domain-specific heuristic problem-solving program which has synthetic organic chemistry as its domain of expertise. SYNCHEM is able to discover synthesis routes for complex and chemically interesting molecules without guidance on the part of its user. Its main components are a chemist-oriented user interface (called KIS) for problem submission, output interpretation and knowledge base maintenance and refinement, a self-guided problem-space search module (the inference engine), a system maintenance module (called HUG), and the knowledge base itself.
Despite the availability of a highly-refined user interface to the knowledge base that provides the chemist with easily understood graphical access to the reaction library, and even though a new reaction schema of moderate complexity can be entered into the library and validated for both syntactic and semantic correctness in an hour or so, it has been very difficult to engage domain experts to the long-term dedication and intensity of commitment necessary to create a production-quality knowledge base. We are currently addressing this problem via a research program in inductive and deductive machine learning.
As the knowledge base grows larger, we will be forced to confront the inevitable inflation in the costs entailed by the substantial increase in both the selection and expressiveness of applicable operators that will have to be examined at each node in the combinatorially expanded region of the search space to be explored. We expect to meet the need for additional computing horsepower by exploiting the inherent suitability of the heuristic search paradigm for distributed/parallel execution. In particular, we plan to adopt the network Linda uncoupled process approach for distribution of the AND/OR problem-space search over the CS department's CER multiprocessor workstation network.