A novel function–structure hierarchical network construction method for crowdsourced product concept design

Abstract Crowdsourced product concept design draws on diverse creative inputs; however, the resulting textual proposals often vary widely in writing style, semantic organization, and functional logic. This heterogeneity poses major challenges for unified knowledge modeling and systematic integration. To address this problem, this study proposes a Function-Structure Hierarchical Network (FSHN) construction method that combines semantic parsing with Bayesian updating to achieve a consistent representation of heterogeneous design knowledge. First, expert-defined functional weights are derived using the Analytic Hierarchy Process (AHP) and encoded as a Dirichlet prior distribution. Next, an automated natural language processing (NLP) pipeline extracts functional and structural entities from design texts, while a fine-tuned BERT model performs multi-label functional semantic mapping. A Dirichlet-multinomial Bayesian updating mechanism is then applied to integrate expert priors with empirical textual evidence, allowing functional weights to be dynamically refined. Experiments on real-world crowdsourced design cases show that the proposed method effectively integrates heterogeneous textual knowledge in a structured manner. The updated functional weights capture both expert intent and data-driven evidence, and the resulting FSHN offers a semantically coherent and interpretable representation. Overall, the proposed framework provides a robust basis for comparing design proposals, supports knowledge reuse, and offers methodological insights into systematically improving the quality of crowdsourced product concept design.