Hi, I’m still looking into this but wanted to share some preliminary findings. Unfortunately we don’t have any user-friendly documentation about what normalization we do for the JSON output. As I look through the codebase, it looks to me like ISSNs are the only metadata field that we currently normalize.
However, we are in the process of standing up a new internal data model at Crossref, and many more metadata fields are planned for normalization. Here is the issue for that work, and here are the tests/specs. Be advised that this work is still pending and the specification is subject to change.