It came from Kenyan annotators for ChatGPT.

The next dataset can come from ChatGPT users.