Dynamic system explanation: DySE, a framework that evolves to reason about complex systems - lessons learned

The large amount of knowledge contained in the scientific literature can be mined using natural language processing and utilized to automatically construct models of complex networks in order to obtain a greater understanding of complex systems. In this paper, we describe the Dynamic System Explanation (DySE) framework, which configures hybrid models and executes simulations over time, relying on a granular computing approach and a range of different element update functions. A standardized tabular format assembles the collected knowledge into networks for parameterization. The Discrete Stochastic Heterogeneous (DiSH) simulator outputs trajectories of state changes for all model elements, thus providing a means for running thousands of in silico scenarios in seconds. Trajectories are also analyzed using statistical model checking to verify against known or desired system properties, determined from text or numerical data. DySE can automatically extend models when additional knowledge is available, and model extension is integrated with model checking to test the validity of additional interactions and dynamics, and enable iterative model updating and improvements. Here, we focus in particular on the difficulties and complications that arise when attempting to automate and integrate information extraction with model assembly and analysis. We discuss how these obstacles are addressed in DySE, the lessons learned from developing and using DySE, and plans for future developments.