Aggregation in order to get an average score (for example) would have to be done on the client's side, but for structuring them based on https://github.com/nostr-protocol/nips/pull/532 I would do something like:

{"tags": [["r", "https://www.themoviedb.org/movie/197-braveheart"], ["L", "review"], ["l", "movie", "review"], ["quality", 0.9]], content: "I like freedom"}

This might be displayed as a 4.5 star review for Braveheart, searchable by namespace ("review") and label within that namespace ("movie").