How to match up similar strings?

Question

Hello all,
I have two tables that have supplier names in them. Some names are similar, but not perfect, so I can't do a name = name command. Here is the code that will help you understand what I'm asking about.
create table abc (supp_cd varchar2(50),
                  supp_nm varchar2(50));
create table xyz (supp_cd varchar2(50),
                  supp_nm varchar2(50));
insert into abc values ('12345', 'KROGER');
insert into abc values ('23456', 'WALMART');
insert into abc values ('34567', 'KMART');
insert into abc values ('45678', 'BEST BUY');
insert into xyz values ('12345', 'KROGER INC');
insert into xyz values ('23456', 'THE WALMART CO');
insert into xyz values ('34567', 'KMART');
insert into xyz values ('45678', 'BST BUY');
select abc.supp_cd as abc_supp_cd, abc.supp_nm as abc_supp_nm, xyz.supp_cd as xyz_supp_cd, xyz.supp_nm as xyz_supp_nm
from abc join xyz on abc.supp_cd = xyz.supp_cd
where abc.supp_nm like xyz.supp_nm;Right now, my code is giving me only:
ABC_SUPP_CD         |ABC_SUPP_NM         |XYZ_SUPP_CD         |XYZ_SUPP_NM         |
34567               |KMART               |34567               |KMART               |When I want:
ABC_SUPP_CD         |ABC_SUPP_NM         |XYZ_SUPP_CD         |XYZ_SUPP_NM         |
12345               |KROGER              |12345               |KROGER INC          |
23456               |WALMART             |23456               |THE WALMART CO      |
34567               |KMART               |34567               |KMART               |
45678               |BEST BUY            |45678               |BST BUY             |I'm not even sure if this is possible, but can someone steer me in the right direction to get similar names that match up?
Thanks ahead of time guys.
PS. I'm running Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

Answer 1

Hi,
For a similar problem, I wrote once a function that returned a number between 0 and 100, reflecting how much 2 strings had in common (100 meant the strings were identical, 0 meant they didn't seem similar at all). Given such a function, you can match rows where the anmes were similar like this:
FROM    abc
JOIN      xyz  ON   abc.supp_cd  = xyz.supp_cd
          AND  nm_similiar (abc.supp_nm, xyz.supp_nm)  >= 70The nm_similar function would start by standardizing the 2 given names, e.g. converting to all capital letters, removing punctuation, and removing words such as 'THE' that you want to ignore. You may not want to ignore words like 'CO', but you probably do want to standardize them, so that 'COMP', 'CMPNY', 'CPY' and 'COMPANY' all turn into 'CO'. I made a separate function for standardization. I also corrected some common mispellings in the standardizaion function. For example, one company was actually named 'HUSQVARNA', but, more often than not, people added a 'U' right after the 'Q'.
Once you've got the names standardized, you can use functions from utl_match to see how similr they are.