File S1 - Contraction of Online Response to Major Events

Supporting Information. Figure S1, Characteristic of data set 1 with a one hour timestep. Figure S2, Characteristics of data set 2a: Forum thread during the presidential election night. Figure S3, Characteristics of data set 2b: Forum thread one week before the presidential election night. Figure S4, Characteristics of data set 3: Enron email corpus. Figure S5, Characteristics of data set 4a: Twitter snow storm. Figure S6, Characteristics of data set 4b: Facebook snow storm. Figure S7, Characteristics of data set 5: App.net. Figure S8, Length distribution of all the tweets in data set 1, and corresponding fits of various distributions. Figure S9, Length distribution of all the posts in data set 2a, and corresponding fits of various distributions. Figure S10, Detail from Fig. S9 showing the tail of the distribution. Figure S11, Dependence of the lognormal fitted parameters with the volume. Figure S12, Analysis of the `ideal' subset of tweets. Figure S13, Characteristic of data set 1 with a 5 min timestep. Figure S14, Characteristic of data set 1 with a 3 hours timestep. Table S1, Best fit MSEs () for the tested distributions for the Twitter data set 1. Table S2, Best fit MSEs () for the tested distributions for the forum data set 2a. Table S3, Characteristic parameters measured on data set 1 for different timesteps.
(PDF)

补充信息。图S1：时间步长为1小时的数据集1的特征；图S2：数据集2a的特征——总统选举之夜的论坛讨论串；图S3：数据集2b的特征——总统选举之夜一周前的论坛讨论串；图S4：数据集3的特征：安然邮件语料库（Enron email corpus）；图S5：数据集4a的特征：推特（Twitter）暴风雪数据集；图S6：数据集4b的特征：脸书（Facebook）暴风雪数据集；图S7：数据集5的特征：App.net；图S8：数据集1中所有推文的长度分布及各类分布的对应拟合结果；图S9：数据集2a中所有帖子的长度分布及各类分布的对应拟合结果；图S10：图S9的细节，展示分布的尾部特征；图S11：对数正态分布拟合参数随数据量的变化关系；图S12：推文“理想”子集的分析；图S13：时间步长为5分钟的数据集1的特征；图S14：时间步长为3小时的数据集1的特征；表S1：针对推特数据集1的各测试分布的最优拟合均方误差（MSE, Mean Squared Error）；表S2：针对论坛数据集2a的各测试分布的最优拟合均方误差；表S3：不同时间步长下数据集1的实测特征参数。（PDF）