I have a huge corpus of documents that are filled with acronyms. It is mostly government stuff. Currently we use regex to translates, but the regex performs poorly and requires a lot of manual fixing. I haven’t been able to google this question (it just brings up lists of machine learning/deep learning acronyms).