Building Large Models of Law with NomosT

Laws and regulations impact the design of software systems, as they introduce new requirements and constraint existing ones. The analysis of a software system and the degree to which it complies with applicable laws can be greatly facilitated by models of applicable laws. However, laws are inherently voluminous, often consisting of hundreds of pages of text, and so are their models, consisting of thousands of concepts and relationships. This paper studies the possibility of building models of law semi-automatically by using the NomosT tool. Specifically, we present the NomosT architecture and the process by which a user constructs a model of law semi-automatically, by first annotating the text of a law and then generating from it a model. We then evaluate the performance of the tool relative to building a model of a piece of law manually. In addition, we offer statistics on the quality of the final output that suggest that tool supported generation of models of law reduces substantially human effort without affecting the quality of the output.