The original (des-moines-register.json) file size is 56.4MB.
After removing the superfluous nesting, the cleaned-up JSON (des-moines-register-clean.json) comes out to 29.5MB, or 52.3% of its original size.
In des-moines-register-abbrev.json, I also tried abbreviating the n-gram keys, e.g. instead of "1 gram", naming the key "1". But surprisingly this only reduced the file size by like 400 bytes?
The last thing I tried was removing all grams with only 1 occurrence. Google ngram does this as well, much more drastically:
we only consider ngrams that occur in at least 40 books. Otherwise the dataset would balloon in size
— via their info page
You can find this file at des-moines-register-trim.json. The file size goes down to 1.4MB, or 2.4% of its original size and 4.7% of the cleaned-up file size.