(A) study on parallelism for deductive retrieval by using dataflow concept

Abstract

In this thesis, a deduction scheme which exploits parallelism for the application of deductive question-answering is proposed. In the deduction scheme, the inference mechanism is separated from that of the retrieval operations. In the inference, full parallelism is exploited by using the dataflow concept. In the retrieval of facts, AND-OR parallelism is realized by a relational database machine. Based on the deduction scheme, an experimental deduction system is simulated, in which Hier-archical Dataflow Database Machine (HRDM)(1) is adopted as the underlying database machine.