Address Matching Using REGEXP_LIKE

Question

Hey all,
I am a newbie to REGEXP and have an issue where I have 2 spatial tables, for each object in table1 and am doing an SDO_NN to find the nearest 2 objects in table2. I also have an address column for each spatial table. I am trying to compare address1 (from table1) to both addresses (address2 from table2) to find the most likely match candidate.
After reading about REGEXP it seemed to me the REGEXP_LIKE was most likely best way of achieving my matches.
What I have come up with works alot of times but not always and I don't quite understand it.
Here is an example where I don't understand what is going on where I used REGEXP_LIKE as this
REGEXP_LIKE(address1,replace(address2,' ','|'))
SELECT
  id, 
  address1,
  address2,
  CASE
    WHEN address1=address2 THEN 'EXACT MATCH'
    WHEN REGEXP_LIKE(address1,replace(address2,' ','|')) THEN 'PROBABLE MATCH'
    ELSE 'NON MATCH'
  END AS TEST
FROM
  SELECT
    1 ID, '930 OLD STEESE HWY' address1, '930 OLD STEESE HWY' address2
  FROM
    DUAL 
  UNION SELECT
    3 ID, '930 OLD STEESE HWY' address1, '920 OLD STEESE HWY' address2
  FROM
    DUAL   
  UNION SELECT
    3 ID, '930 OLD STEESE HWY' address1, '99 COLLEGE RD' address2
  FROM
    DUAL   
  UNION SELECT
    4 ID, '930 OLD STEESE HWY' address1, '80 5TH ST' address2
  FROM
    DUAL     
RESULTS
ID ADDRESS1            ADDRESS2            TEST
1  930 OLD STEESE HWY  930 OLD STEESE HWY  EXACT MATCH
2  930 OLD STEESE HWY  920 OLD STEESE HWY  PROBABLE MATCH
3  930 OLD STEESE HWY  99 COLLEGE RD       NON MATCH
4  930 OLD STEESE HWY  80 5TH ST           PROBABLE MATCHI am really confused as to the 4th line in the results?
Any thoughts and or suggestions?
Appreciate it!
Cheers,
Eric

Answer 1

Second and fourth row match because there is a word in address2 that is present (at least as substring) in address1: For the fourth row ST is substring of STEESE.
If you want a PROBALE MATCH only when there's a word in address2 that is present as a word in address1, you can change this way:
SQL> SELECT
  2    id,
  3    address1,
  4    address2,
  5    CASE
  6      WHEN address1=address2 THEN 'EXACT MATCH'
  7      WHEN REGEXP_LIKE(address1,'(^| )'||replace(address2,' ','|')||'($| )') THEN 'PROBABLE MATCH'
  8      ELSE 'NON MATCH'
  9    END AS TEST
10  FROM
11    (
12    SELECT
13      1 ID, '930 OLD STEESE HWY' address1, '930 OLD STEESE HWY' address2
14    FROM
15      DUAL
16    UNION SELECT
17      3 ID, '930 OLD STEESE HWY' address1, '920 OLD STEESE HWY' address2
18    FROM
19      DUAL
20    UNION SELECT
21      3 ID, '930 OLD STEESE HWY' address1, '99 COLLEGE RD' address2
22    FROM
23      DUAL
24    UNION SELECT
25      4 ID, '930 OLD STEESE HWY' address1, '80 5TH ST' address2
26    FROM
27      DUAL
28    );
        ID ADDRESS1           ADDRESS2           TEST
         1 930 OLD STEESE HWY 930 OLD STEESE HWY EXACT MATCH
         3 930 OLD STEESE HWY 920 OLD STEESE HWY PROBABLE MATCH
         3 930 OLD STEESE HWY 99 COLLEGE RD      NON MATCH
         4 930 OLD STEESE HWY 80 5TH ST          NON MATCHIn Oracle 11g you can use REGEXP_COUNT to count the common words between address1 and address2:
  2    id,
  3    address1,
  4    address2,
  5    CASE
  6      WHEN address1=address2 THEN 'EXACT MATCH'
  7      WHEN REGEXP_LIKE(address1,'(^| )'||replace(address2,' ','|')||'($| )') THEN 'PROBABLE MATCH'
  8      ELSE 'NON MATCH'
  9    END AS TEST,
10    regexp_count(address1,'(^| )'||replace(address2,' ','|')||'($| )') common_words
11  FROM
12    (
13    SELECT
14      1 ID, '930 OLD STEESE HWY' address1, '930 OLD STEESE HWY' address2
15    FROM
16      DUAL
17    UNION SELECT
18      3 ID, '930 OLD STEESE HWY' address1, '920 OLD STEESE HWY' address2
19    FROM
20      DUAL
21    UNION SELECT
22      3 ID, '930 OLD STEESE HWY' address1, '99 COLLEGE RD' address2
23    FROM
24      DUAL
25    UNION SELECT
26      4 ID, '930 OLD STEESE HWY' address1, '80 5TH ST' address2
27    FROM
28      DUAL
29    );
        ID ADDRESS1           ADDRESS2           TEST           COMMON_WORDS
         1 930 OLD STEESE HWY 930 OLD STEESE HWY EXACT MATCH               4
         3 930 OLD STEESE HWY 920 OLD STEESE HWY PROBABLE MATCH            3
         3 930 OLD STEESE HWY 99 COLLEGE RD      NON MATCH                 0
         4 930 OLD STEESE HWY 80 5TH ST          NON MATCH                 0Max
[My Italian Oracle blog|http://oracleitalia.wordpress.com/2009/12/29/estrarre-i-dati-in-formato-xml-da-sql/]
Edited by: Massimo Ruocchio on Dec 29, 2009 10:48 PM
Added regexp_count example.