Towards a computable policy tool capable of leveraging domain knowledge in knowledge graphs

Abstract

Policies are sets of rules that describe suitable responses that correspond to some given set of conditions. They are crucial tools when it comes to the decision-making process, regardless of domain or profession. Policies are typically conveyed through the use of natural language. Policies represented in natural language can, at times, be subject to ambiguity and differing interpretations. Moreover, natural language policies do not necessarily provide the domain knowledge or context necessary for a practitioner to derive a correct conclusion. By contrast, computable policies can be used to guarantee desirable traits such as standardization, machine-readability, and automatic policy evaluation through the use of frameworks. Most computable policy frameworks are domain-specific. Using frameworks beyond their intended domain(s) can require considerable amounts of configuration and the changes required can sometimes result in issues. One way to address this problem is to use domain information expressed as knowledge graphs within the computable policies themselves. Doing this would provide machine-readability, mitigate terminology inconsistency, reduce ambiguity, and include domain knowledge inside the policy evaluation process. The majority of frameworks do not provide this capacity, and those that do typically lack tools to help work within them. For these reasons we propose ADAPT: a domain-agnostic policy tool that leverages domain knowledge stored using W3C recommended standards in ontologies and provenance on the web. Our solution enables the construction, visualization, and management of computable policies that integrate machine-understandable domain knowledge into the policy evaluation process. The proposed solution aims to provide a platform for developing semantically-enabled computable policies while reducing the amount of background knowledge in Semantic Web technology that is necessary to create them.