Then you should take a look at https://github.com/nostr-protocol/nips/pull/879

My implementation is different, since I believe a rating is generic (like a reaction) and can be applied generically.

I use m-tags to differentiate between different kinds of classes. Default is event, but you can rate whatever you want with a single event kind.

This linked NIP argues for different rating kinds for different classes.

I personally think it sucks for discoverability, but it may work.