There's automated solutions out there, and free providers like the GDELT project and ACLED that go way beyond sentiment analysis.
They're great, but what if you're tasked with solving the problem by yourself?
I gave it a shot and found that Google RSS + Newspaper3k + Zero-Shot model gets you surprisingly far in classifying hundreds of articles.
I'd love to hear your thoughts. How would you solve the problem, and what features would you add?