Bridging Semantic Knowledge and Generative AI: A Modular Framework for Automated Reporting in Digital Commissioning Management
This paper proposes a framework integrating semantic triple stores with large language models (LLMs) to enhance automated report generation in digital commissioning systems. It addresses the challenge of efficiently extracting and analysing complex industrial data by combining RDF graphs with LLM-ba...
| Autores: | , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2026 |
| País: | España |
| Institución: | Varias* (Consorci de Biblioteques Universitáries de Catalunya, Centre de Serveis Científics i Acadèmics de Catalunya) |
| Repositorio: | Recercat. Dipósit de la Recerca de Catalunya |
| OAI Identifier: | oai:recercat.cat:2445/227879 |
| Acceso en línea: | https://hdl.handle.net/2445/227879 |
| Access Level: | acceso embargado |
| Palabra clave: | Gestió de projectes Intel·ligència artificial Web semàntic Project management Artificial intelligence Semantic web |
| Sumario: | This paper proposes a framework integrating semantic triple stores with large language models (LLMs) to enhance automated report generation in digital commissioning systems. It addresses the challenge of efficiently extracting and analysing complex industrial data by combining RDF graphs with LLM-based natural language processing. The approach involves: 1) developing an ontology for digital commissioning; 2) structuring data as RDF triples; 3) integrating triplestores with LLMs using LangGraph and LangChain. This enables natural language querying with high semantic accuracy. Using a simulated dataset, the system achieved 100% accuracy in SPARQL query generation across diverse question types, effectively handling entity relationships, hierarchies, and query complexities. The framework bridges structured data and natural language interfaces in industrial contexts, improving efficiency and accuracy in data retrieval and reporting. Future research should explore scalability, heterogeneous datasets, and data quality challenges in real-world implementations. |
|---|