Architecture and Prototype System for Automatically Processing Natural-Language Statements of Policy

摘要

Organizations are policy-driven entities. Policy bases can be very large and complex; these factors are compounded by the dynamic nature of policy evolution. Thus, comprehension of the ramifications of both policy modification and assurance of the consistency, completeness, and correctness of a policy base necessarily requires some level of computer-based support. A policy workbench is an integrated set of computer-based tools for developing, reasoning about, and maintaining policy. A workbench takes as input a computationally equivalent form of policy statements. In this thesis we explore approaches for translating natural-language policy statements into their equivalent computational form with minimal user interaction. We present the architecture of a natural-language input-processing tool (NLIPT), which we designed to augment a policy workbench. NLIPT components consist of an extractor, index-term generator, structural modeler, and logic modeler. We experimented with a prototype of the extractor. The extractor successfully parsed twenty-seven of a sample of ninety-nine of U. S. Department of Defense security policy statements. An additional twenty-one statements were correctly parsed based on the syntactic structure of the input.