Declarative Merging of and Reasoning about Decision Diagrams

Decision diagrams (DDs) are a popular means for decision making, e.g., in clinical guidelines. Some applications require to integrate multiple related yet different diagrams into a single one, for which algorithms have been developed. However, existing merging tools are monolithic, application-tailored programs with no clear interface to the actual merging procedures, which makes their reuse hard if not impossible. We present a general, declarative framework for merging and manipulating decision diagram tasks based on a belief set merging framework. Its modular architecture hides details of the merging algorithm and supports preand user-defined merging operators, which can be flexibly arranged in merging plans to express complex merging tasks. Changing and restructuring merging tasks becomes easy, and relieves the user from (repetitive) manual integration to focus on experimenting with different merging strategies, which is vital for applications, as discussed for an example from DNA classification. Our framework supports also reasoning over DDs using answer set programming (ASP), which allows to drive the merging process and select results based on the application needs.