construct urban knowledge graph

This work proposes an end-to-end approach to efficiently transform fully modular CityGML 2.0 datasets into compact urban knowledge graphs in Neo4j, addressing the limitations of CityGML’s XML-based exchange format, which is semantically rich yet ill-suited for direct querying or analytical tasks. The method innovatively integrates R-tree spatial indexing with a model-agnostic text-to-Cypher translation mechanism, enabling natural language-driven joint semantic and spatial queries. The system combines a large language model (LLM) interface, a graph database backend, and OGC 3D Tiles for visualization, and demonstrates its feasibility on Hamburg’s LoD2 dataset through complex use cases such as identifying suitable rooftops for greening. The architecture ensures data locality, auditability of results, and mitigation of LLM hallucinations.

Urban modeling has long neglected citizens’ cognitive processes and social interactions. Method: This paper proposes a novel synergistic framework integrating multimodal generative AI with agent-based modeling (ABM), establishing the paradigm of “Social AI in Urban Ecosystems.” It combines multimodal large language models, geographic-semantic embeddings, cross-scale knowledge fusion, and large-scale text/image knowledge extraction to achieve rich, semantically grounded representations of citizens’ mental models, behavioral patterns, and social interactions. Contribution/Results: The work transcends conventional physically oriented urban modeling by introducing a computationally tractable and dynamically simulatable socio-spatial coupled model. This advances urban planning’s decision-support capabilities—particularly in human-centered design, systemic resilience, and sustainability—by enabling rigorous, evidence-informed, and socially aware simulation and policy evaluation.

This paper addresses critical challenges in urban network modeling—inefficient data acquisition, insufficient multimodal integration, and limited openness—by proposing a systematic solution built upon OSMnx. Methodologically, it extends OSMnx to enable automated downloading and joint modeling of multimodal transportation networks (e.g., walking, cycling, bus), integrates high-fidelity geometric parsing, dynamic attribute embedding, and graph-theoretic spatial analysis, and establishes a reproducible, extensible open-science framework. Contributions include: (1) the first deep integration of open science principles into urban computing infrastructure design; (2) substantial improvements in street-network modeling accuracy and cross-scale analytical capability; and (3) robust support for interdisciplinary research in geography, transportation engineering, and computer science. The resulting tool has become a mainstream infrastructure in open urban analytics, adopted in hundreds of empirical studies worldwide.

To address the challenges of acquiring generalizable knowledge and adapting to downstream tasks in urban region representation, this paper proposes a spatial-entity-augmented regional graph modeling framework. It pioneers the unification of graph pretraining and graph prompting in urban computing: (i) a subgraph-centered self-supervised pretraining paradigm is designed, integrating structure-aware masked subgraph reconstruction with subgraph-level contrastive learning; and (ii) a dual-path graph prompting mechanism is introduced, jointly injecting task-specific knowledge via explicit template-based and implicit soft prompts. Evaluated across multiple cities and diverse downstream tasks—including POI prediction, crowd flow forecasting, and functional zone identification—the method consistently outperforms state-of-the-art approaches, achieving average accuracy improvements of 5.2%–9.7%. The framework significantly enhances the generalizability of regional representations and enables effective cross-regional knowledge transfer.

Large language models (LLMs) struggle to effectively comprehend graph-structured data due to their inherent sequence-based architecture and lack of native graph-aware representations. Method: This paper introduces *graph laws*—statistically derived, topologically parameterized features that are interpretable as natural language descriptions—establishing a novel paradigm for representing graphs as LLM-compatible inputs. We systematically construct a multi-dimensional graph law framework spanning macro/micro scales, low/high orders, and static/dynamic properties, integrating graph-theoretic analysis, multi-scale observational modeling, and natural language alignment techniques, while establishing semantic mappings to downstream graph tasks and retrieval-augmented generation (RAG) scenarios. Results: Experiments demonstrate that graph laws substantially mitigate LLM hallucination, overcome context-length limitations, and enable end-to-end graph reasoning. The approach achieves strong generalization across diverse domains, including molecular design, recommender systems, and protein structure modeling.

This study addresses the challenge faced by non-expert users and interdisciplinary researchers in efficiently accessing semantically rich yet structurally complex 3D urban models formatted in specialized standards. To bridge this gap, the authors propose the first lightweight and extensible conversational query framework that integrates large language models—such as GPT-OSS, Gemini 3.1, and GPT-5.4—with spatial and graph databases to automatically translate natural language queries into structured database commands and generate corresponding visualizations. The framework supports four types of complex queries: spatial, graph-based, cross-database, and multi-turn dialogues. Evaluated on 54 test cases, it achieves answer accuracy rates of 85.2%–100%, visualization correctness of 92.9%–100%, and successfully completes all queries with an average of fewer than three retries.

This work proposes a method to construct universal global location representations using only OpenStreetMap (OSM) data, enabling diverse geospatial applications without reliance on remote sensing imagery. By modeling geographic environments as heterogeneous graphs incorporating roads, buildings, land use, and points of interest, the approach integrates a multi-scale graph encoder with spherical harmonic positional encoding. Notably, it introduces a CLIP-style contrastive learning framework to OSM graph data for the first time, leveraging solely map topology and semantic information to generate high-quality location embeddings. Evaluated across seven downstream tasks spanning climate, ecology, socioeconomic, and public health domains, the method achieves strong performance—significantly outperforming satellite-image baselines in socioeconomic and public health tasks—demonstrating OSM’s remarkable capacity to encode human activity patterns.

This study addresses the lack of a unified knowledge representation framework in current urban digital twins, which hinders multi-source data fusion, provenance tracking, spatial reasoning, and machine learning. The authors propose, for the first time, a provenance-aware urban knowledge graph that integrates CityGML and OpenStreetMap data, incorporates roof material prediction and LoD3 geometric reconstruction, and employs confidence-weighted edges to align cross-source entities while preserving a complete evidence chain. The resulting 180 GiB knowledge graph comprises 180 million nodes and 220 million edges. The work also introduces novel benchmark tasks to evaluate capabilities in 3D spatial reasoning, cross-source consistency judgment, and detection of unsatisfiable queries. Experiments reveal that commercial large language models achieve only 54–69% accuracy, while open-source 7B models perform as low as 6–19% and fail entirely to recognize unsolvable queries.

This work addresses the need for unified and efficient knowledge provisioning in large language models by proposing a novel architecture that integrates relational and property graph data models. The approach leverages record addresses from log files as immutable reference values in place of traditional foreign keys, enabling efficient graph-style link traversal instead of costly join queries while natively supporting triple-based knowledge representation. The resulting unified knowledge service framework combines the structural rigor of relational models with the flexible associative capabilities of graph models, significantly enhancing knowledge retrieval efficiency and effectively supporting knowledge integration and invocation in generative AI systems.

This work presents the first systematic approach to instance-free schema inference under property graph query transformations. Given a ProGS input schema and a G-CORE query, the authors propose a multi-layer mapping technique that translates property graphs, schemas, and queries into RDF, SHACL, and SPARQL CONSTRUCT representations, respectively, enabling automatic derivation of structural constraints on the output graph via description logic reasoning. By leveraging RDF reification and cross-language semantic bridging, the method establishes a sound and semantically equivalent metatheoretical foundation. This enables generic output schema inference applicable to any input graph conforming to the given schema, while formally verifying both the correctness of the derived constraints and the semantic fidelity of the mappings.