nostr:nprofile1qqsdcnxssmxheed3sv4d7n7azggj3xyq6tr799dukrngfsq6emnhcpspz4mhxue69uhhyetvv9ujuerpd46hxtnfduhszxmhwden5te0w35x2en0wfjhxapwdehhxarjxyhxxmmd9uq3camnwvaz7tm5dpjkx6t5v9jx2mpwdehhxarjxyhxxmmd9u2756e3 could embeddings help solve the spam problem?

I'm thinking a relay could run classification on incoming events and automatically add the embedding events to its database.  Clients could choose to use those embeddings to detect "spam-like substances" and auto-hide or flag posts for user review.

The relay could even auto-delete the most spammy events based on an analysis of the embeddings.

Is that reasonable, or too computationally expensive?