Thanks to the amazing efforts of one of our data scientists, we have created a website to track toxicity on #Gab in real-time. The link is observatory.pushshift.io
This site will allow you to see new Gab content as it is posted to Gab.
#datascience #bigdata #dataviz #hatespeech
Conversation
Replying to
This is great. Some have "(flagged as: {racist,anti-semitic})". How is that determined? Is that another model or algorithm?
1
Replying to
Right now we are using the Perspective API to assign toxicity scores and for classification. We're currently in the early stages of this project and evaluating different models to gauge their effectiveness and accuracy. Given time and additional tweaks, the accuracy will improve.
1
1
3
Replying to
What determines when it's labeled as "racist" vs. "anti-semiic" vs. no explicit label, though?
1
Replying to
Where exactly are you seeing these labels? Do you have a screen shot? The developer leading this project is unavailable at the moment but I'll definitely ask and get more details for you. If you have a screen shot of where you see these labels, that will help as well. Thanks!
1
Replying to
Thanks! Here's an example with a two posts that include the "flagged" message:
1
1
Replying to
Ahh thanks for the example. I'll find out more info for you when the developer is around tomorrow. Thanks for the screen shot!