ABSTRACT
This PhD thesis addresses the multiple challenges in knowledge graph construction and validation, based on real-world data from engineering domains. It presents a comprehensive methodology for Linked Data generation, demonstrated through the creation of a city-scale building energy benchmarking knowledge graph for Barcelona city. This work also introduces a novel benchmark, ERA-SHACL-Benchmark, for evaluating SHACL engines with real railway infrastructure data, revealing crucial performance and conformance discrepancies. Furthermore, it proposes an end-to-end visual Single Source of Truth (SSoT) framework, to involve domain experts in the knowledge graph lifecycle, leveraging Chowlk notation, and adopted in a real use case: the emissions inventory data space. Collectively, this research bridges the gap between theoretical Semantic Web advancements and their practical application in engineering, contributing in validated methodologies, valuable datasets, rigorous benchmarks, and user-centric frameworks for reliable and maintainable knowledge-based systems.
PhD Advisors:
CANDIDATE
Edgar Alexis Martínez Sarmiento is a PhD candidate in Computer Engineering and Information Technologies at BEE Group, CIMNE’s Innovation Unit in Building, Energy and Environment. His research focuses on Linked Data generation for buildings and open government data, ontology development, and statistical learning applied to time-series data.