The DIBBS blackboard control architecture and its application to distributed natural language processing

Abstract

The authors present DIBBS, a domain-independent blackboard system that has been used for applications in distributed natural language processing (NLP). It is shown how the explicit representation of control knowledge as units on the DIBBS control blackboard allows an application to implement dynamic control strategies and reason about its own problem-solving performance. The DIBBS truth maintenance dependency network allows robust treatment of control issues in conflict resolution. All of these properties makes DIBBS especially appropriate for applications in distributed NLP, such as the DIOGENES and DIANA systems.<>