How Does It Work?
Using Machine Learning To Protect Your Feed
At it’s core WatchTower uses the Twitter API to pull Tweets and analyze them. If a tweet is seen to contain ‘unfavorable’ content (i.e. hate speech, extremism, or misinformation) then WatchTower will review the tweet owner’s last 50 tweets and creates a confidence score for how likely that user is to be an author of such content. Once this process has concluded WatchTower saves this information into a database which is used when a user authenticates to the platform and all users will be blocked based on the provided confidence threshold.
WatchTower was build as an entry for the Chirp 2022 Developer Challenge, and serves as a wrapper around the Twitter API and several machine learning frameworks. The source code for WatchTower is open source and can be found on HuggingFace. At current, WatchTower only supports the identification of Extremism (with misinformation and hate speech planned for a future 2.0 release). The extremism identification is performed using a tool called Pinpoint, which is also open source.
- Block hate speech, misinformation, and extremism.
- Tailor your own personal thresholds for what’s blocked.
- Use it all completely free!
“Gain control of your feed, by blocking the content you don’t want to see.”