Why this site is a knowledge graph
Most portfolios are flat lists of projects and papers. This one is an RDF knowledge graph rendered as an interactive force-directed layout. Here's why.
The problem with flat portfolios
A typical academic portfolio is a set of disconnected lists: publications here, projects there, a bio somewhere else. The structure says nothing about how these things relate. A reader who wants to understand what connects your work has to do the synthesis themselves.
That bothered me. My research is fundamentally about making connections explicit — building knowledge graphs that let people and machines traverse relationships between heterogeneous data. It felt wrong to present that work as a flat page.
What this site actually does
The landing page of this site loads two Turtle files — an ontology and a dataset — and parses them client-side with N3.js. The parsed triples become nodes and edges in a D3.js force-directed graph.
Every entity on the graph — a research project, a skill, a publication, a site section — is a real RDF resource with a URI. The edges are real RDF predicates: pp:worksOn, pp:usesTechnology, schema:about. Node size reflects connectivity: the more things an entity connects to, the bigger it appears.
Section nodes are clickable. Click “Research” in the graph and you navigate to the research page. The graph is the navigation.
Why RDF, not just JSON
I could have modelled the graph as a JSON file with a nodes and edges array. It would have been simpler. But RDF gives me something JSON doesn’t: a schema.
The ontology defines what kinds of things exist (skills, interests, projects, publications) and what relationships are valid between them. When I add a new entity, the structure tells me what properties it should have. That’s exactly the value proposition of knowledge graphs in any domain — and if I’m going to argue for that value in my research, the least I can do is practice it myself.
The tech stack
- Data: Turtle files served as static assets, no server needed
- Parsing: N3.js in the browser
- Visualisation: D3.js force simulation with SVG rendering
- Framework: Astro, with the graph as a client-side island
- Hosting: GitHub Pages
The entire site is static. The RDF is just files. There’s no SPARQL endpoint (yet) — though the data is structured enough that adding one with Comunica is a matter of importing the library.
What I skipped
Plenty. There’s no reasoning, no inference, no federated queries. The ontology is minimal and personal — it’s not trying to be reusable beyond this site. The graph doesn’t update dynamically; I edit the Turtle files and redeploy.
These are deliberate choices. The point isn’t to showcase every feature of the semantic web stack. The point is that structured, connected data — even at this small scale — tells a richer story than flat lists.
Eat your own dog food
If your research is about making knowledge more connected and explorable, your own portfolio should demonstrate that. This site is a small, opinionated argument for knowledge graphs as a way of thinking about information — not just in cultural heritage institutions, but anywhere you have things and relationships between them.