Influence Distribution

Influence in a text network can be measured using betweenness centrality for each node. It shows, how often a certain concept appears between any two randomly chosen concepts from a text. The more often it links the concepts from different topical clusters together, the more influential the concept is. 

In InfraNodus, we use Jenks natural breaks clusterization algorithm (a simplified version of CKMeans) to determine the number of top influence nodes that have a high measure of betweenness centrality, different from the rest of the networks.

However, it is also important to see how well this influence is distributed among the different topical clusters in a network.

For instance, if all the top nodes in a network are concentrated in the same topic, then we can say that the network is not very diverse. A single cluster takes all the influence, even if it's distributed among different concepts.

Alternatively, if each of the 4 top influential nodes belongs to a distinct community, we can say that influence in this network is equally distributed among the different groups, so there is more pluralism in a text network. 

Influence distribution is shown in the Network Structure panel of InfraNodus: