This paper presents the Digital Humanities Hub at the University of Barcelona, an institutional initiative designed to consolidate dispersed research databases, strengthen public-facing digital infrastructures, and foster engaged scholarship. The Hub responds to a structural problem widely recognised in the humanities: while project-level datasets are abundant, specialised, and often publicly funded, they frequently remain fragmented, inconsistently documented, and at risk of long-term loss. Many originate from short-term projects, theses, or individual initiatives lacking sustainable maintenance, resulting in valuable knowledge being locked in silos and inaccessible to wider communities.

The Hub addresses this challenge by developing an interoperable, open, and socially oriented infrastructure grounded in three pillars: (1) standardisation and FAIR/LOD compliance, (2) knowledge-graph integration supported by neurosymbolic AI, and (3) participatory, human-centered design. Together, these elements aim to create an environment where datasets can be connected, explored, enhanced, and reused, enabling new research possibilities while expanding public engagement with humanities knowledge.

The first pillar focuses on harmonising data practices across the institution. Building on European standards and open data methodologies, Hub proposes a phased model for improving data quality and interoperability: inventory and localisation of existing resources; refinement and normalisation to address errors, inconsistencies, and divergent formats; and the adoption of FAIR and Linked Open Data principles. Through the use of ontologies, controlled vocabularies, and persistent identifiers, the infrastructure ensures semantic coherence and facilitates long-term accessibility via RDF- and SPARQL-based technologies. This systematic approach supports institutional goals by preventing the abandonment of smaller but culturally valuable datasets and by promoting sustainable, connected stewardship of humanities data.

The second pillar integrates knowledge graphs and neurosymbolic AI to enable reliable and transparent information retrieval. A multi-layered knowledge graph connects heterogeneous institutional databases with external resources such as Wikidata, supporting both human-readable exploration and machine reasoning. A Retrieval-Augmented Generation (RAG) layer, grounded in the curated graph, provides explainable generative outputs and mitigates hallucinations and biases typical of large language models. This hybrid approach transforms isolated datasets into interconnected semantic ecosystems, enabling more nuanced and relational analyses that reflect the complex structures of cultural and historical knowledge.

The third pillar centres on engaged, participatory methodologies. Researchers, students, librarians, Wikimedia contributors, and community stakeholders take part in co-design processes that shape the ontology, user interfaces, and interpretative layers of the graph. This human-centered dimension ensures accessibility, relevance, and ethical representation, especially important for datasets that document sensitive histories or historically marginalised groups. Participation is understood not only as a methodological choice but as a commitment to inclusive knowledge creation and responsible digital humanities practice.

Beyond technical innovation, the Hub foregrounds the institutional and societal value of digital research infrastructures. By adopting social-return-on-investment perspectives, it frames humanities data as a public good with significant cultural, educational, and civic impact. Transforming scattered datasets into a connected, participatory knowledge infrastructure demonstrates how universities can align open data practices, AI technologies, and public engagement to strengthen their role as democratic knowledge institutions.

This contribution argues that hybrid infrastructures are essential for advancing socially responsive, open, and resilient digital humanities.