Knowledge-driven generative AI for museum product innovation: A BTM-AIGC-TOPSIS framework

Amidst the digital transformation of industrial design, extracting actionable knowledge from massive user-generated content (UGC) and maintaining controllability in Artificial Intelligence-Generated Content (AIGC) remain significant challenges for product innovation. This study proposes a novel data-driven intelligent design framework that systematically bridges the gap between unstructured consumer feedback and creative solution generation. The methodology follows a rigorous three-stage information-processing loop. First, for knowledge discovery, the Biterm Topic Model (BTM) is employed to mitigate semantic sparsity in 114,877 online reviews, followed by Grounded Theory to distill 18 multidimensional user demand elements. Second, to achieve knowledge transformation, the Analytic Hierarchy Process (AHP) quantifies requirement weights, which are subsequently engineered into structured semantic prompts for generative design engines. This stage ensures that generative creativity is effectively constrained by empirical user data. Finally, for decision optimization, an Entropy-Weighted TOPSIS model is constructed to objectively evaluate the proposals. Applying this framework to a case study of museum products, the results demonstrate that the proposed method significantly enhances design efficiency and market alignment. This research contributes a reusable computational design paradigm for the intelligent product development lifecycle, offering significant implications for industrial digital transformation in the AI era.