Ask HN: Improving Data Quality for Addresses

I am currently working on improving data quality of global postal addresses that we store. It has turned out to be a lot harder problem to solve than I had originally imagined owing to no standardized format for addresses. I am leaning towards using ML to solve it but cant figure out a way to do that. Can forum members share their experiences if they have solved such problems in the past. I would also appreciate being pointed to some ML/DL research papers/ techniques talking about similar stuff