FIG. 1: Network of Twitter users who share New York Times online articles. Links are "follow" relationships between the users. The network layout is obtained by pulling followers and followees close to each other, while pushing apart unconnected nodes. The long range links that are visible occur where relatively few links connect groups of nodes that are otherwise pushed far apart. (main image) The same network layout with user nodes colored according to the topic of articles they share most. Links are shaded with the color of the follower.
FIG. 2: Network of Twitter users who share New York Times online articles. Links are “follow” relationships between the users. The network layout is obtained by pulling followers and followees close to each other, while pushing apart unconnected nodes. The long range links that are visible occur where relatively few links connect groups of nodes that are otherwise pushed far apart. (main image) The same network layout with user nodes colored according to the topic of articles they share most. Links are shaded with the color of the follower.
FIG. 3: The same network structure given in Figure 1, links are omitted for clarity. For each of the sub-communities (A, B, Upper A, Lower A, Upper B, and Lower B) the top three URL topic distributions are identified and these categories are given in the bar charts. The width of each bar-chart component is proportional to the number of users in the corresponding category. The total width of each bar chart is proportional to the total number of users in each community.
FIG. 4: Comparison between clusters based on topic, location and biography attributes. Each vertical axis shows the most dominant terms associated with that cluster. The top figure compares Cluster A (left) and Cluster B (right), as well as the upper and lower domains shown by distinct colors. The bottom left two axes and right two axes compare upper A to lower A, and upper B to lower B, respectively.
FIG. 5: The network structure layout given in Figure 1 with edges hidden and users colored according to their geographical locations. For each cluster, the distribution of the locations are given in bar charts. The width of each bar-chart component is proportional to the number of users in the corresponding category. The total width of each bar chart is proportional to the total number of users in each cluster.