Figure 1.
The number of nodes or edges varies for different interaction threshold
In the following part of the present work, we set to extract a large enough network with convincing interaction strength.
Figure 2.
(Color online) The giant connected cluster of a network sample with
(a) is the network structure, in which each node stands for a user and the link between two users represents the interaction between them. Based on this topology, we color each node by its emotion, i.e., the sentiment with the maximum tweets published by this node in the sampling period. In (b), the red stands for anger, the green represents joy, the blue stands for sadness and the black represents disgust. The regions of same color indicate that closely connected nodes share the same sentiment.
Figure 3.
Correlations with error-bar for different emotions as the hop distance varies.
Large means a pair of users are far away from each other in the social network we build. Here
is fixed.
Figure 4.
The emotion sequence is randomly shuffled to test the correlation significance.
Figure 5.
Pearson correlations of different for different networks extracted by varying
The case of is not considered here because of the weak sentiment correlation found in Figure 3.
Figure 6.
Here is fixed to 10 to reduce the data sparsity.
Because the network is relatively small, the largest degree we get is only 30. Therefore, the results in Figure 6a just demonstrate that when the degree is small, how the sentiments' correlations vary with node degrees. While regarding to Figure 6b, the linear bin is used to get emotion sequences for nodes with clusterings within the same bin.
Figure 7.
The example Chinese keywords extracted for anger(left) and sadness(right), respectively.
The top 20 keywords are also translated into English, which could be found through http://goo.gl/tl4q45.