An Innovative Approach to the Semantic Integration of Construction and Sustainability Information Using Graphrag

Sustainable construction necessitates a holistic view of buildings over their entire lifecycle. This work introduces an innovative, data-driven approach for the semantic integration of construction and sustainability information. An ontology was developed and implemented as a hierarchically structured knowledge graph within a Neo4j database. Heterogeneous, predominantly unstructured data sources – from technical datasheets and Environmental Product Declarations (EPDs) to safety data sheets and certificates – are automatically processed using large language models (LLMs) and incorporated into the graph. By integrating GraphRAG (Graph-Retrieval Augmented Generation), fact-based responses derived from the graph are generated via LLMs. The ontology was populated with 35 construction product datasets, yielding a knowledge graph with 3,186 nodes and 3,262 edges. Validation using 21 representative competency questions – covering technical, economic, ecological, and sociocultural aspects – demonstrated complete resolution of technical queries, though economic inquiries revealed gaps due to missing pricing or cost data. Nonetheless, the results underscore the potential of this approach, significantly advancing semantic data models in the construction and sustainability sectors. Ultimately, the fusion of knowledge graphs with LLMs paves the way for fact-based, AI-assisted decision-making in construction.