Fig 1.
Overview of the Spiegel DAG (top), and one part enlarged (bottom).
The DAG is broken into two lines in the top figure to fit the whole graph in the available width. On the bottom figure, dashed links indicate descendants of the tag “Germany” which are not shown.
Fig 2.
Part of the Guardian’s “Environment” branch, in the component “World news”.
Hierarchical levels are separated by dashed lines.
Fig 3.
Similarities between the news portals’ DAGs, according to the mutual information-based linearized information similarity measure described in Similarity of hierarchies.
Fig 4.
The Spiegel (top) and the Guardian’s (bottom) reduced DAG structures, providing the largest overall similarity in our analysis.
For clarity, Spiegel’s DAG is broken into two lines. Background colors show the result obtained by applying the similarity measure given in Eq 1 to the given branch and the most similar branch from the other hierarchy. Note that sub-branches on all hierarchical levels have their own color.
Fig 5.
Guardian’s “Technology” branch (top) and New York Times’ “Computers and the Internet” component (bottom).
Hierarchical levels are separated by dashed lines. Grey tags do not appear in both DAGs, however, they connect branches containing common tags.
Table 1.
Ratios of inversions between centralities and real-world occurrence frequencies, calculated for tag pairs coappearing in statistically significant numbers.
Totally random case corresponds to 0.5.
Table 2.
Characteristic level showing the highest level of an idealized hierarchy to which an average connected component corresponds.
Table 3.
Ratios of giant and dwarf branches among all branches, size-weighted.