Fuzzy search not returning results?

Question

I'm executing a phonetic search on the nm_resource column. my application allows a call center employee to search on the resource name (nm_resource), if the resource is not found then they will enter a new one. The problem is someone may have already entered the resource name but spelled it incorrectly resulting in duplicate records for the same resource name. To enable the call center to retrieve records that may have the same sound but are spelled differently we have implemented the fuzzy search capability of Oracle text. Things have been going very nicely for the most part with the exception of this one issue we're trying to understand.
Using the query below we're searching for the resource name "rosies" the actual record in the database was entered as "rosy's". the search returns (rosies,rosie's,rosys) and does not return ---> rosy's <--- the record i'm interested in
it is reasonable to expect rosy's to be returned in the result set? my query should retunn the max fuzzy expansions and all fuzzy scores.
select score(1), nm_resource, ADDR_RSRC_ST_LN_1, id_resource, ADDR_RSRC_CITY FROM caps_resource where
CONTAINS (nm_resource,'fuzzy(rosies, 0, 5000, weight)',1)>0
union
select /*+index(caps_resource ind_caps_resource_8)*/ 10, nm_resource, ADDR_RSRC_ST_LN_1, id_resource, ADDR_RSRC_CITY from caps_resource 
where NM_RESOURCE_UPPER like upper(replace(replace('%' || 'rosies' || '%',' '), '-'))
and rownum<500 order by 1 DESC;
any help explaining this is much appriciated.
Regards,

Answer 1

When you index "Rosy's", by default it sees the apostropohe as a delimiter and tokenizes and indexes "Rosy" and "s" separately. So, you could only find it by searching the singular form or the singular form obtained by using stemming. However, if you set the apostrophoe as a skipjoin, then it tokenizes and indexes "Rosys" as one token that you can then search for that using "rosies":. Please see the demonstration below. You might also be interested in soundex, which can be used with Oracle Text, or the functions in the utl_match package or metaphone.
SCOTT@orcl_11g> CREATE TABLE caps_resource
  2    (nm_resource  VARCHAR2 (30))
  3  /
Table created.
SCOTT@orcl_11g> INSERT ALL
  2  INTO caps_resource VALUES ('Rosy''s')
  3  SELECT * FROM DUAL
  4  /
1 row created.
SCOTT@orcl_11g> SELECT * FROM caps_resource
  2  /
NM_RESOURCE
Rosy's
SCOTT@orcl_11g> CREATE INDEX your_text_idx ON caps_resource (nm_resource)
  2  INDEXTYPE IS CTXSYS.CONTEXT
  3  PARAMETERS
  4       ('STOPLIST CTXSYS.EMPTY_STOPLIST')
  5  /
Index created.
SCOTT@orcl_11g> SELECT token_text FROM dr$your_text_idx$i
  2  /
TOKEN_TEXT
ROSY
S
SCOTT@orcl_11g> SELECT * FROM caps_resource
  2  WHERE  CONTAINS (nm_resource, 'FUZZY (rosies, 0, 5000, weight)') > 0
  3  /
no rows selected
SCOTT@orcl_11g> DROP INDEX your_text_idx
  2  /
Index dropped.
SCOTT@orcl_11g> BEGIN
  2    CTX_DDL.CREATE_PREFERENCE ('your_lexer', 'BASIC_LEXER');
  3    CTX_DDL.SET_ATTRIBUTE ('your_lexer', 'SKIPJOINS', '''');
  4  END;
  5  /
PL/SQL procedure successfully completed.
SCOTT@orcl_11g> CREATE INDEX your_text_idx ON caps_resource (nm_resource)
  2  INDEXTYPE IS CTXSYS.CONTEXT
  3  PARAMETERS
  4    ('STOPLIST CTXSYS.EMPTY_STOPLIST
  5        LEXER       your_lexer')
  6  /
Index created.
SCOTT@orcl_11g> SELECT token_text FROM dr$your_text_idx$i
  2  /
TOKEN_TEXT
ROSYS
SCOTT@orcl_11g> SELECT * FROM caps_resource
  2  WHERE  CONTAINS (nm_resource, 'FUZZY (rosies, 0, 5000, weight)') > 0
  3  /
NM_RESOURCE
Rosy's
SCOTT@orcl_11g>