To automate design, a computational representation of all the nuances of design is necessary. This has been challenging because even humans are usually unsure of the best ways to abstract design into a formal representation.
For my PhD, I have introduced an automatic, grammar-based system for structuring information (DCC'16). We applied this tool to data from design, architecture, and abstract systems (AI EDAM forthcoming). We used this representation technique for detecting subtle differences between data sets, as accurately as the existing state of the art, but requiring almost 10 times less data.