Revealing Matrices
Go to Figure 1 2 3 4 5 6 7 8

Select figure

Read summary

See credits

Buy the book

Open access pdf

Visit author's website

Tweet about this

Share on Facebook

Revealing Matrices / Figure 6

Collapsing subdivided entries in the raw data uncovers interesting complex features

Click on the figure to open a larger image [and use browser to go back]
Looking at the raw CENSUS data, Documents, Monuments and Bibliographic Citations are too fragmented to draw meaningful quantitative conclusions.
a) Single parts of larger Documents are connected to single Monument nodes;
b) Single parts of larger Documents are held together by ‘parent links‘ forming a forest of Document trees in the Document–Document cell (cf. Figure 5);
c) In the raw Monument–Document cell the single parts of larger Documents are separate. Surprisingly there is nevertheless a fairly large connected component of overlapping document parts, i.e. text paragraphs and overview figures that are connected to multiple Monuments – forming a small hairball. However, the structure of the connected component looks rather random, without many interesting features.

If we reduce trees of large Documents, Monuments, and bibliographic Citations to single nodes, a transformation takes place, and more sophisticated complex structure in the database becomes apparent.
a‘) After the tree collapse, the entire Document is now represented as a single node, connected to the monuments via weighted links. The link weight corresponds to the collapsed number of links between two collapsed nodes;
b‘) The collapsed Document, Monument and Bibliographic Citation trees are now represented as weighted nodes;
c‘) In the reduced Monument–Document cell, 89.9% of all collapsed nodes are connected in a single ‘Giant Connected Component (GCC)‘. Brush like structures in the diagram indicate a very large number of Monuments, characterized by a lack of sufficient information, subsequently connected to only a single Document. Architectural Monuments, easily recognizable by their large node weight due to their subdivisions, cluster together as their coverage overlaps across documents, for example across guide books and city maps.

Download a pdf version of the current figure
April 15, 2010 version
Creative Commons License
All visualizations on the Revealing Matrices website by Maximilian Schich
are licensed under a Creative Commons Attribution-Noncommercial 3.0 Germany License.
Permissions beyond the scope of this license may be available at