Mapping consumer voice into engineering insight: a structured language model-driven design support framework for electric vehicles

Consumer-generated narratives contain rich experiential knowledge for product design, yet their perceptual and weakly structured nature limits direct translation into actionable engineering insight. This challenge is particularly evident in EV design, where user insight plays a critical role but remains weakly grounded in formal engineering representations. To address this gap, this study proposes a Structured consumer semantic insight small language model (SCSI-SLM). First, Structured Semantic Encoding (SSE) maps weakly structured user narratives into a multi-level engineering semantic space, transforming perceptual expressions into comparable and reusable semantic tokens. Second, Product-User Dynamic Mapping (PUDM) decouples product-centric performance assessment from user-centric preference representation, enabling importance-performance analysis and preference heterogeneity to be modelled from the same corpus. Third, an Engineering Design Knowledge Graph (EDKG) integrates user profiles, product attributes, perceptual features, and textual evidence into an interpretable reasoning space. Building on this representation, a hybrid retrieval-reasoning engine combines semantic similarity search with graph-constrained inference to generate evidence-backed and engineer-readable insight chains. A case study in the EV domain demonstrates that the proposed framework can generate traceable design insights from large-scale consumer feedback, suggesting structured language modelling as a viable paradigm for data-driven engineering design.