Spam Dataset

[Jack Chakany]

Hey Will,

It's become pretty obvious that we need an ml classifier for spam detection, relays and maybe clients too.

You said on Nostr that relay.damus.io has 90GB of data, this is a massive amount and creates a large classification problem.

Something that could be developed is a website that people can go to and get zapped every time they classify a note as spam or not correctly. A note could be shown to multiple people, and if they are consistent than they get zapped a small amount.

I have a 24-core epyc with 1TB of storage and 128GB of ram that i'm renting from a friend of mine, it could be trained on that.

Looking to hear your opinion on this.

- Jack Chakany