About a year ago, I tried visualising the blog content…

…but that was more of a failure.

Now I’ve given it another go. This time I’ve used the Triple Store from the last post and the result is certainly impressive:

It may take a moment for the graph to load and calculate.

Contains data from Wikidata, licensed under CC0.

For mobile devices, available as a PDF download; the file is not updated when new posts are added.

Explanations

The red rectangles represent the individual blog posts.
The yellow diamonds represent blog tags.
Blue circles are Wikidata entities (which also serve as tags).
Green circles represent entities on the page (e.g. linked articles or other metadata entries).

Some objects have coloured borders, which indicate the type. However, only a few types are currently configured.

Pre-selection

The following query is used to visualise the data for the graph above:

PREFIX schema: <http://schema.org/>
SELECT ?s ?p ?o ?isTagged WHERE {
  <https://christianmahnke.de/post/> schema:blogPost ?post .
  ?post ?p ?o .
  BIND(?post AS ?s)
  FILTER(?p NOT IN (
    schema:author,
    schema:url,
    schema:workTranslation,
    <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
  ))
  FILTER(?o NOT IN (
    schema:BlogPosting
  ))
  OPTIONAL {
    ?post schema:identifier ?ident .
    ?ident a schema:PropertyValue ;
            schema:propertyID "projektemacher" ;
            schema:value "tag" .
    BIND(true AS ?isTagged)
  }
}

Implementation

In addition to the combination of HDT and OxiGraph mentioned in the last post, Cytoscape is used for visualisation. The PDF export was created using cytoscape-svg.

Why?

Sure, the graph looks nice, but it also has other advantages:

It allows navigation outside the confines of this blog.
The outer rings used for classification provide insight into the hierarchy of object relationships (“is a”). For example, the Dötlingen Artists’ Colony (Q1797167) is neither a place, nor a group, nor an organisation.
It can help with quality control of tagging; for example, it highlights that libjxl (Q99738405), i.e. the implementation, and JPEG XL (Q72885392), i.e. the file format, have so far been assigned inconsistently.

Further improvements

There is still room for improvement:

The visualisation of the base classes (e.g. people, places, organisations, software) is not yet complete.
The thematic proximity of the nodes could be utilised further in the graph layout.
In general, the layout could be improved.
Cytoscape offers visually appealing visualisations that are more attractive than the one shown above.

Update 22 April 2026

As suggested by Konrad Höffner, there is now a full-screen mode and Cytoscape now uses WebGL.

Christian Mahnke

Blog metadata visualised