I'd treat them as a sorting problem, at least that way, it's easier to see candidates are grouped together. If we can assume every record has a zip code, then you can sort first by zip, and then within the zip, sort by street number and street name (leaving out the str., blvd). Once that's done, it will give you a better idea how to proceed, i.e., what kind of ambiguities people use, and probably do some mapping of the street", "blvd", etc.