I keep uploading texts (daily) that are about 30 pages long on average (~100'000 characters). Since I need to match the text with future texts, but don't want to load the whole content of the document into a database (which would also be too big), I tried to break it down to a 256 character string.

The goal is to match the text with the match future texts, and output the match in %. To make it easier to check, it would be best for me if I can store a 256-character long string on the DB, and generate future texts also into a 256-character long string, and then compare these strings with each other and return the percentage of match. (Similar to this Website)

Here I read that this doesn't make sense to convert these into hashes, as it takes too much processing power. Are there other possibilities? Or what would make sense for my situation?