The data analyzed for this article includes videos, metadata, and social media posts collected from Twitter, YouTube, and Pomegranate Cloud between February 18 and June 2. We downloaded it. over 5,000 videos were posted to these platforms between January 23 (the first day of the campaign video following the State Department’s January 19 genocide declaration) and May 31. On the Pomegranate Cloud, We collected clips targeting Mr. Pompeo by searching for posts that mentioned him after January 19 that contained the video. We have collected the forced labor rejection videos in Xinjiang’s cotton industry from a dedicated section in the app.
On Twitter and YouTube, campaign videos are collected from what we call “stock accounts,” videos whose videos are shared by a coordinated network of more than 300 Twitter accounts. The network appears to have worked in concert to like and repost content supporting Chinese government policies, such as campaign videos, as well as articles and editorials from the media. state communication. To identify networks, we manually define a subset of accounts and their automation metrics, namely posts containing identical content, followed by random strings of characters . We then identify additional network accounts by programmatically looking for other accounts that promote the same content and have similar metrics.
We’ve cataloged over 3,000 unique campaign videos out of more than 5,000 collected. To accurately identify duplicates between videos containing different compression ratios, visual artifacts, and subtitle languages, we calculated a fingerprint for each video by running a frame sample of that video through through Google Cloud Vision image tagging. We have identified videos with similar fingerprints and duration as duplicates. We take samples manually and review the results from this process to minimize false positives and false negatives.
To identify non-campaign videos on YouTube and Twitter, we first get their subtitles using optical character recognition on frames taken from them at regular intervals. We consider videos that do not mention Mr. Pompeo or cotton as videos unrelated to the campaign. Videos on Twitter and YouTube are always subtitled in English and Chinese. Videos from Pomegranate Cloud are only subtitled in Chinese when Uyghur is spoken. We consider all videos collected from Pomegranate Cloud as campaign videos, but do not review each video manually. We also use the captions of YouTube and Twitter videos to analyze their content.