Zexuan Shi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zexuan Shi has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Yunhao Zhang
Department of Visual Design, Hanyang University, Seoul 03063, Republic of Korea
1 shared publication - Zuowen Bao
Department of Design, Sungkyunkwan University, Seoul 03063, Republic of Korea
1 shared publication
Latest Publications
- Open Access
- 8 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 1015-1022
A Weight-Driven Generative AI Framework for Tujia Xilankapu Pattern Design: Integrating AHP and TOPSIS for Computational Design Evaluation
This study proposes a weight-driven generative artificial intelligence (GenAI) framework for the synthesis and computational evaluation of traditional indigenous patterns, utilizing Tujia Xilankapu as the primary paradigm. To mitigate the inherent tendency of GenAI to generate morphologically plausible but culturally superficial outputs, this research operationalizes tacit heritage knowledge into computable parameters via the Analytic Hierarchy Process (AHP). Twelve design indicators are hierarchically structured across three rigorously defined dimensions: cultural semantics, morphological syntax, and visual-chromatic expression. Expert pairwise comparisons yielded eigenvector-based priorities, which are subsequently embedded into a structured prompt-engineering pipeline. High-weight indicators—specifically structural regularity, geometric abstraction, and cultural symbolism—are translated into explicit algorithmic constraints governing AI-driven pattern synthesis. Four generated design permutations are then evaluated using the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). Analytical results indicate that morphological syntax commands the highest criteria weight, while the TOPSIS validation confirms that Case 1 achieves maximum proximity to the idealized cultural-design matrix. This research contributes to computational design by establishing a reproducible, closed-loop workflow that synergizes quantitative expert knowledge modeling, parameterized generative constraints, and proximity-based algorithmic validation. Ultimately, the proposed framework fundamentally enhances the interpretability, structural controllability, and cultural fidelity of GenAI in heritage translation tasks.
- 8 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 1015-1022
Open AccessA Weight-Driven Generative AI Framework for Tujia Xilankapu Pattern Design: Integrating AHP and TOPSIS for Computational Design Evaluation
This study proposes a weight-driven generative artificial intelligence (GenAI) framework for the synthesis and computational evaluation of traditional indigenous patterns, utilizing Tujia Xilankapu as the primary paradigm. To mitigate the inherent tendency of GenAI to generate morphologically plausible but culturally superficial outputs, this research operationalizes tacit heritage knowledge into computable parameters via the Analytic Hierarchy Process (AHP). Twelve design indicators are hierarchically structured across three rigorously defined dimensions: cultural semantics, morphological syntax, and visual-chromatic expression. Expert pairwise comparisons yielded eigenvector-based priorities, which are subsequently embedded into a structured prompt-engineering pipeline. High-weight indicators—specifically structural regularity, geometric abstraction, and cultural symbolism—are translated into explicit algorithmic constraints governing AI-driven pattern synthesis. Four generated design permutations are then evaluated using the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). Analytical results indicate that morphological syntax commands the highest criteria weight, while the TOPSIS validation confirms that Case 1 achieves maximum proximity to the idealized cultural-design matrix. This research contributes to computational design by establishing a reproducible, closed-loop workflow that synergizes quantitative expert knowledge modeling, parameterized generative constraints, and proximity-based algorithmic validation. Ultimately, the proposed framework fundamentally enhances the interpretability, structural controllability, and cultural fidelity of GenAI in heritage translation tasks.