Abstract:
This work provides a systematic analysis of cultural heritage (CH) semantic portals as a representative application domain for knowledge graphs and Semantic Web technologies. The paper presents a PRISMA-based systematic literature review of 77 academic papers published in the last 10 years describing 68 cultural heritage semantic portals through five key research dimensions. First, we examine the semantic layer, documenting the adoption of ontologies (CIDOC-CRM, EDM) and controlled vocabularies (Getty AAT, GeoNames). Second, we investigate resource integration workflows, from data acquisition methods to preprocessing formats. Third, we catalog presentation paradigms across web, mobile, and immersive (AR/VR) platforms, while evaluating framework usage (Omeka S, ResearchSpace) and the presence of advanced research features. Fourth, we examine the technical systems used, such as database tools and ontology software. Finally, we look at organizational factors including project motivation, participating institutions, and how portals are evaluated. Our findings demonstrate uneven adoption of Semantic Web standards, with 20% of portals using non-interoperable ontologies and 41% lacking vocabulary links. While human-centric interfaces dominate, only 24% support machine-readable access via SPARQL. Furthermore, Over 50% of the surveyed portals are inaccessible—either due to missing links in the publications or because the provided URLs lead to broken/non-existent pages. The study highlights the need for: (1) improved documentation of implementation practices, (2) reusable open-source tools for data harmonization, and (3) strategies to bridge the gap between large-scale and smaller CH projects. By identifying common practices and persistent challenges, this work contributes not only to the development of interoperable CH applications but also to the broader understanding of challenges in designing and deploying knowledge graph-based systems. Future work will investigate the publication of the collected survey results as a structured, semantically described dataset.
