For addresses, I'd recommend first using nested REPLACE()'s (SQL Server 2000, at least, will easily go 100 or 120 deep) to standardize abbreviations. You'll have to look at your own data to include common misspellings, as well.
Personally, I also standardize on the singular abbreviations, on the assumption that plural vs. singular when most other factors are the same is more often a typo than two different streets with the same <entity> having the same numeric address on both.
Splitting addresses into street #, street name, and suite can be very... interesting.