Name ordering is clear, so what about the clustering and similarity views?

**Cluster view:** The newspapers with the same color on the diagonal is a cluster.
For example *(posta, hurriyet, milliyet)* is a cluster and *(yenisafak, Haber7, AkitGazetem)* is another one.
So, our methodology seems to be doing pretty well, huh?
The clustering algorithm measures how well a network is decomposed into modular communities.
Our graph G=(V,E) is a weighted clique where set of V is the 35 newspapers and W(E) is the pairwise similarities.
11 clusters visualized in this matrix are generated based on modularity measure.

**Similarity view:** As the newspapers appear more to the right/down on the axis, they are (cumulatively) less similar to others,
i.e. followers of these newspapers are isolated, or these newspapers may be called as marginals.

More on this project: Identifying Turkish News Media Groups with Twitter.