EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Research on Material Translation Matching Model for Regional Environmental Design Based on NLP and Cultural Gene Mapping

Jing Zhang

International Scientific Technical and Economic Research · 11.08.2026 · 4 (3) · 60-93

DOI: 10.67541/istaer2626 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Scientific Technical and Economic Research, online erschienen am 11.08.2026. Verfasst von Jing Zhang. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The selection of materials for regional environment design has long been trapped in the cross-modal gap between cultural semantics and physical properties, and existing methods find it difficult to transform the tacit knowledge in the construction tradition into a computable matching basis. This paper proposes a material translation matching model based on natural language processing and cultural gene mapping. Through deep semantic extraction of multi-source heterogeneous corpora, a heterogeneous information network with "geography, history and technology" as hyperedges is constructed to realize the joint embedding representation of material nodes and cultural symbol nodes. Aiming at the fuzziness and dynamics of design intention, a multi-dimensional constraint tensor modeling method based on fuzzy membership function is designed, and Pareto front ranking and context-aware relaxation coefficient are introduced to realize online adaptive adjustment of elastic matching window. At the matching computing level, a dynamic graph attention aggregation network is proposed to capture semantic associations in heterogeneous structures through node-level, relationship -level and path level attention mechanism, and integrate cross-modal contrast learning to align text constraint space and graph embedding space. Experiments are carried out on the standardized test set covering three regions, namely, water towns in the south of the Yangtze River, red brick in the south of Fujian, and forest pan in the west of Sichuan. The matching accuracy @1 of the complete model reaches 66.8%, which is 22.8 percentage points higher than the optimal baseline model. The lightweight variant achieves a single‑inference latency of 42.8 ms at the Jetson Xavier NX edge with an accuracy loss of only 6.0 percentage points. In the blind evaluation, 78.5% of the recommended results were judged as highly consistent by senior architects, and the relative deviation of physical and chemical indicators was within the engineering allowable ranges. Ablation experiments and generalization tests verify the effectiveness of each module and the ability of cross-regional migration of the model.

Zitieren

APA

Zhang, J. (2026). Research on Material Translation Matching Model for Regional Environmental Design Based on NLP and Cultural Gene Mapping. International Scientific Technical and Economic Research, 4(3), 60-93. https://doi.org/10.67541/istaer2626

Vancouver

Zhang J. Research on Material Translation Matching Model for Regional Environmental Design Based on NLP and Cultural Gene Mapping. International Scientific Technical and Economic Research. 2026;4(3):60-93. doi:10.67541/istaer2626

BibTeX
@article{Zhang2026,
  author = {Zhang, Jing},
  title = {Research on Material Translation Matching Model for Regional Environmental Design Based on NLP and Cultural Gene Mapping},
  journal = {International Scientific Technical and Economic Research},
  year = {2026},
  volume = {4},
  number = {3},
  pages = {60--93},
  doi = {10.67541/istaer2626}
}
Kostenlos starten →