For additional context, a database is said to be more normalized (or in higher normal form) when more of the entities in it can be looked up by a unique identifier in their own table.
Like if your users table had a bookmarks column/field that held a list of urls. Normalizing that data would involve creating a bookmarks table/collection, creating an entry for each unique url across all users in said table, and then updating the bookmarks columns to hold bookmark ids instead of urls directly.
The ids are usually automatically created by the database but can be anything unique. The benefits are deduplication of data, faster lookups, and much more efficient reasoning about the dataset as a whole. (Think about how you’d check if anyone has bookmarked a given url before and after the normalization.)
This was way longer than I intended, I hope it helps.