Ahhh, excellent suggestion; I believe the master database has a soundex column.
You're getting some great advice. On the do-it-yourself front, take a look at the SOUNDEX routine built into SQL. Helps a lot on dealing with minor spelling variations.
Someone else mentioned regex in 10 (we're on 9, but we can move the data-set to an XE instance). Not searching Oracle directlly, I did find this:
quote: The sample data consists of a list of address strings. The table that holds this data is named XXADDR and contains only one column, ADDRESS_LINE1, of type VARCHAR2(300). Our test scenario involves a requirement to standardize street abbreviations to comply with the U.S. Postal Service standard (described in Appendix E of Postal Addressing Standards). A sample of the standard is shown below: -- http://www.freelists.org/webpage/oracle-l