报告:中国互联网审查效率之高令人咋舌
Just How Fast Are China's Internet Censors? Very, a New Study Finds
中国社交媒体的审查人员从来都不睡觉，不过在夜深人静的时候，他们的工作效率也会打些折扣。
China's social media censors never sleep, though they do fall behind on their work late at night.
这是一份新研究报告的若干发现之一，报告作者为独立研究员祝涛(音)和几位研究中国人气颇高的新浪微博SINA+1.80%审查方式的美国学者。
That's one of a number of findings from a new study by independent researcher Zhu Tao and a handful of U.S. academics that analyzes the mechanics of censorship on China's most popular microblogging platform, Sina Corp.'s Weibo.
通过对2012年7月至9月期间那些曾违反审查规定的用户在微博上发布的238万个帖子进行分析研究，报告列出了很可能为新浪“编辑”团队采用的审查方式。这个团队是全球监控社交媒体最大规模行动的核心力量。
Based on an analysis of 2.38 million Weibo posts published between July and September 2012 by users known to have run afoul of censors, the report lays out the approach most likely taken by Sina's team of 'editors' who are at the heart of the world's largest effort to control social media.
虽然背审查黑锅的一般来说都是中国宣传部门，但实际上，真正对网络内容“下手”的是中国国内互联网公司本身，而中国政府只是在它认为这些公司没有做好这一工作时才采取行动。
Though Chinese propaganda authorities typically shoulder the blame for censorship, in practice the country's Internet companies police their own content, with the government only taking action when it doesn't think the companies have done a proper job.
作为中国社交媒体网站中最活跃、政治气氛最浓厚的平台，新浪微博在源源不断地输出敏感帖子。让很多正在研究中国政府审查实践的学者印象深刻的是，新浪审查团队管控这些帖子时速度和效率都高得惊人。
As the most active and political of China's social media sites, Weibo hosts a constant torrent of sensitive posts. What impresses many academics, who are increasingly studying the Chinese government's censorship practices, is how startlingly fast and efficient Sina's teams of censors are in managing that flow.
报告说：在我们的数据集中，有5%的帖子在出来后的8分钟内被删除，近30%的帖子在30分钟内被删除，考虑到微博处理的数据集规模之大，这个速度实在太快了，特别是在5到10分钟的高峰时段，特别是考虑到这项工作无法以全自动的方式来完成。
'In our data set, 5% of the deletions happened in the first 8 minutes, and within 30 minutes, almost 30% of the deletions were finished＃.considering the big data set that Weibo has to process, the speed, especially at the 5 to 10 minutes peak, is fast, especially considering it cannot be processed in a fully automated way,' the authors write.
报告说，微博上平均每分钟有7万条帖子发布，这种情况下，审查人员很可能会使用含有敏感词的不同列表，同时关注那些很可能会说出敏感词语的用户，这样就能助过滤系统一臂之力。审查人员发现他们认为应该被过滤的词后，他们首先会删除原帖，然后搜索所有转发帖，将其一网打尽，通常五分钟内就能完成工作。
With 70,000 messages posted to Weibo per minute on average, the report says censors probably refer to different lists of sensitive terms and follow posters most likely to say sensitive things to get a leg up on the filtering process. When they find an entry they deem should be filtered, they first delete the post, and then track down all reposts, deleting them in one rapid swoop, usually accomplishing the task in 5 minutes.
被新浪审查团队盯上而删除的帖子中，82%是转发帖。
Of the deleted messages tracked by the team, 82% were reposts of earlier messages
研究人员发现，有10%的被删帖是在发布24小之后被删除的。其中大多数帖子使用了新词或逃避自动关键词过滤的暗语。例如，重庆市前市委书记薄熙来去年倒台丑闻爆出的时候，微博用户为绕开对薄的屏蔽，用“西红柿”指代他，其中“西”与他名字中的“熙”字同音。
The researchers found that 10% of deleted posts are erased after 24 hours. The bulk of those contain neologisms or code words designed to evade automatic keyword filtering. For instance, during the scandal surrounding fallen Chongqing Communist Party boss Bo Xilai last year, microbloggers circumvented blocked searches for Mr. Bo by referring to him as 'tomato,' a pun on one of the characters in his name.
最终这些暗语会被发现且屏蔽，从而导致新暗语的产生，然后再度被屏蔽，如此循环往复。
Eventually the code words get discovered and blocked, which leads to the invention of new code words, which require new blocks, and on and on.
报告作者说，由于审查人员不断发现新暗语，他们会使用关键搜索词来回溯，然后很快删除相关帖子。例如，他们发现有44条不同的微博中使用了新发现的敏感性暗语，然后在区区五分钟内就将其全部删除。
As the censors uncover fresh code words, the authors say, they use keyword searches to go back and delete them quickly. In one instance, they found that 44 independent posts containing a newly discovered sensitive code word were deleted in a matter of five minutes.
若有用户一再公然发布敏感内容，新浪有时会直接将其销户。在报告研究的3.5万个账户中，有300个账户有如此遭遇。
When a user repeatedly and flagrantly posts sensitive content, Sina will sometimes simply delete the account, something that happened to 300 of the 3,5000 accounts tracked in the study.
不过，尽管新浪审查人员警惕性很高，但他们似乎也是要休息的。研究发现，审查工作虽然一天24小时不停歇，但凌晨那几个小时会有一段“风平浪静”的时光。报告说，上午的删帖高峰时段表明，审查人员的进度有所落后，他们一边要翻看前晚的帖子，一边还要处理上午发的新贴，然后到接近中午或午后不久他们就赶上进度了。
Vigilant as they may be, however, Sina's censors appear to take breaks: The study found that censorship happens 24-hours a day, but that there's a lull in the early morning hours. 'The morning-hour spike suggests that the censors are behind in the morning, both catching up on overnight posts and dealing with a fresh influx of posts from morning risers,' the authors write. 'They catch up by late morning or early afternoon.'
对微博用户来说，他们发的帖子越敏感，审查人员上午的压力就越大，或许这一点能令他们感到欣慰。但据去年11月发布的另一份报告统计，大多数时候，审查人员是最终的赢家。这份报告与上述报告出自同样一群作者之手。
It's perhaps some consolation for posters that the more sensitive posts they write, the more stressful a morning it is for the censors. But most of the time censors win in the end, according to another paper published by the same group in November.
较早的那份报告说，审查一开始，某话题的讨论频率会明显降低。报告作者将审查人员比作是白血球。
'Starting from the point where censorship begins, the frequency of the topic drops dramatically,' reads the earlier report (pdf), in which the authors liken the censors to white blood cells.
报告说，生物免疫系统的工作重点不是尽快达到零感染的水平，而是阻止传染物扩散并干扰到生物体的健康；微博的审查重点似乎同样务实，只要敏感话题没有大规模传播，一些敏感帖子在某段时间内是可以被容忍的。
'A biological immune system's focus is not on reaching zero infection as quickly as possible, but rather on stopping infections from spreading and interfering with the health of the organism. Weibo's censorship seems to have a similarly pragmatic focus. As long as sensitive topics are not going 'viral,' a few sensitive posts can be tolerated for some period of time.'
华尔街日报
2013-03-11
华尔街日报
1207次
点赞(0)
收藏
登录
