Evaluating the performance of architectural structures using NLP-enhanced fuzzy logic

Purpose This study presents a hybrid framework integrating natural language processing (NLP) with fuzzy logic to quantitatively assess the spatial and functional performance of architectural structures, addressing the limitations of conventional and expert-dependent evaluation methods. Design/methodology/approach The framework processes textual design descriptions, extracting semantic features and mapping them to fuzzy logic criteria such as development, geometric order, modularity, safety and flexibility. Validation was performed through five spaceframe structure case studies, comparing NLP-generated scores with expert-assigned fuzzy scores to evaluate accuracy and reliability. This study focuses primarily on the formal and spatial performance of architectural structures, particularly space frame and barrel vault typologies. Future work will extend this framework to include functional performance variables such as user interaction, activity mapping and occupancy behavior, following emerging approaches in AI-assisted spatial analytics. Findings The results demonstrate that the NLP-fuzzy logic framework closely reproduces expert evaluations while providing a scalable, automated and objective assessment tool. By converting qualitative architectural narratives into structured quantitative metrics, the model enables consistent, transparent and reliable performance evaluation, supporting informed decision-making in early design stages. Research limitations/implications The primary limitation of this study lies in its reliance on the quality and clarity of architectural text descriptions; ambiguous or poorly structured narratives may affect the accuracy of semantic extraction. Additionally, the framework currently focuses on a predefined set of fuzzy criteria, which may not fully capture all dimensions of architectural performance across diverse typologies. Future work could expand the model’s adaptability by incorporating more advanced language models and broader performance metrics. Nonetheless, the research provides strong implications for reducing subjectivity and enhancing scalability in architectural evaluation using AI-driven methods. Practical implications This framework offers architects, engineers and design evaluators a practical tool to objectively assess spatial and functional performance based on qualitative design narratives. By automating the interpretation of architectural language, it reduces dependency on expert judgment and facilitates faster, more consistent evaluations during early design phases. The method can be integrated into digital design platforms to support real-time feedback, aiding in form selection, structural clarity and performance optimization. Its application to spaceframe structures demonstrates potential for broader use in complex architectural systems, enhancing data-informed decision-making in both educational and professional architectural practice. Social implications By democratizing access to performance evaluation tools, this research supports more inclusive and transparent design processes. The framework empowers designers, especially in resource-limited contexts, to assess architectural quality without requiring constant expert oversight. It promotes equity in design evaluation by providing a standardized, AI-driven method that can be universally applied. Additionally, by facilitating better-informed decisions in early stages, the model contributes to creating safer, more functional and human-centered built environments. This aligns with broader societal goals of sustainability, accessibility and efficiency in architecture, ultimately enhancing the quality of life through improved design outcomes. Originality/value This research introduces a novel methodology linking qualitative architectural language to quantitative assessment, enhancing the application of AI in architectural analysis. The hybrid approach offers a practical, expert-independent tool for designers and establishes a new paradigm for automated, language-based performance evaluation in architecture.