arXiv数据论文发布BuildGraph合成多原型建筑知识图谱公开数据集

View PDF HTML (experimental)查看 PDF HTML（实验性）

Abstract:Semantic querying of building knowledge graphs (KGs) underpins the integration of artificial intelligence into building operations, from natural-language access to cross-building analytics, but such KGs are rarely public owing to proprietary, security, and cost barriers. BuildGraph is a synthetic building KG dataset of 120 buildings in the Brick Schema ontology, grounded in U.S. Department of Energy prototype models and sensor-placement patterns from real buildings. It spans eight commercial building types across three ASHRAE energy-code vintages, with five realizations per archetype varying URI naming, sensor-attachment predicates, and topology. A 75-query SPARQL benchmark confirms structural completeness: BuildGraph reaches 91.5% Query Answerability Rate versus 35.8% for 59 real-world Brick files. Independently, it reproduces real buildings' sensor-type proportions on their shared vocabulary (cosine 0.86-0.94), evidence of realistic instrumentation where measurable. A downstream text-to-SPARQL experiment with Gemma 4 (26B) reaches 27.3% Row-Matching F1 (+17.5 pp over zero-shot) on 12 held-out buildings. BuildGraph gives facility managers and digital-twin developers a testbed for portable analytics and natural-language interfaces, and its dataset, generator, and benchmark are openly available at this https URL. 摘要：建筑知识图谱（KG）的语义查询支撑人工智能融入建筑运营，应用范围从自然语言访问到跨建筑分析；但由于专有、安全和成本障碍，此类 KG 很少公开。BuildGraph 是一个采用 Brick Schema 本体的合成建筑 KG 数据集，包含 120 栋建筑，基于美国能源部原型模型和真实建筑的传感器布置模式构建。它涵盖 3 个 ASHRAE 能源规范年代版本中的 8 种商业建筑类型，每种原型有 5 个实现，在 URI 命名、传感器挂载谓词和拓扑上各不相同。一个包含 75 条查询的 SPARQL 基准证实了结构完整性：BuildGraph 的查询可回答率达到 91.5%，而 59 个真实 Brick 文件仅为 35.8%。独立来看，它在二者共享词汇表上复现了真实建筑的传感器类型比例（余弦值 0.86-0.94），证明在可测量之处具有真实的仪表配置。一项使用 Gemma 4（26B）的下游 text-to-SPARQL 实验在 12 栋留出建筑上达到 27.3% 的 Row-Matching F1（较零样本提高 +17.5 pp）。BuildGraph 为设施经理和数字孪生开发者提供了用于可移植分析和自然语言界面的测试台，其数据集、生成器和基准可在此 https URL 公开获取。

Comments:评论：	22 pages, 3 figures. Data paper. Dataset, generator, and 75-query SPARQL benchmark openly available at this https URL and this https URL22 页，3 幅图。数据论文。数据集、生成器和包含 75 条查询的 SPARQL 基准可在此 https URL 和此 https URL 公开获取。
Subjects:主题：	Databases (cs.DB)数据库（cs.DB）
ACM classes:ACM 分类：	H.2.3; I.2.4; I.2.7
Cite as:引用格式：	arXiv:2610.00038 [cs.DB]
	(or arXiv:2610.00038v1 [cs.DB] for this version) （或本版本使用 arXiv:2610.00038v1 [cs.DB]）
	https://doi.org/10.48550/arXiv.2610.00038 arXiv-issued DOI via DataCite 通过 DataCite 获取 arXiv 签发的 DOI

Submission history提交历史

From: Wooyoung Jung [view email]
[v1] Wed, 2 Sep 2026 21:14:26 UTC (126 KB)
来自：Wooyoung Jung [查看邮箱]
[v1] Wed, 2 Sep 2026 21:14:26 UTC (126 KB)

「英中对照」arXiv数据论文发布BuildGraph合成多原型建筑知识图谱公开数据集

Submission history提交历史