FROM POETRY TO GEOMETRY: A PROCEDURAL AI-BOOSTED APPROACH TO SHAPE GRAMMARS

In the realm of architectural representation and analysis, digital parameterization emerges as a powerful tool for reinterpreting the foundational principles of classical architecture through a contemporary lens. It can be seen as an enabling tool for uncovering the relationship between architecture and geometry. The Mitchell algorithm, better known as shape grammar, is a formal system based on rules that allows for the generation and transformation of geometric shapes in space, through an iterative process that defines a visual and design language. With the goal to retrace the proportional and compositional logics of masters like Palladio, it provides a logical framework capable of translating traditional canons into dynamic and adaptive digital structures (Stiny, Mitchell 1978). This approach leverages hierarchical and ontological management of architectural taxonomy, facilitating the analysis and reproduction of complex historical elements. In the last two decades, the use of visual programming languages (VPL) further enhances the potential of the design process, enabling smooth and iterative parametric modeling. Indeed, these methodologies have been widely implemented in BIM workflows and modeling for years, facilitating the dynamic evolution of 3D systems and enabling models' transparency (Lo Turco et al. 2024). They streamline the integration of design data with models and automate repetitive and complex modeling operations (Battini et al. 2024). Currently, a game changer in all the ICT applications and visualization practices is undoubtedly the generative AI, in particular LLMs can be exploited for converting text in complex scripts and supporting the translation of mathematical proportions in models