Order of words, fuzzy and utl_match

Question

Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Prod
PL/SQL Release 10.2.0.1.0 - Production
"CORE     10.2.0.1.0     Production"
TNS for 32-bit Windows: Version 10.2.0.1.0 - Production
NLSRTL Version 10.2.0.1.0 - Production
create table category(cat_id number(20),cat_type varchar2(3000));
create table category_match(cat_id number(20),cat_type varchar2(3000));
Insert into category (CAT_ID,CAT_TYPE) values (12790,'AUTO CONSULTANTS');
INSERT INTO CATEGORY (CAT_ID,CAT_TYPE) VALUES (23803,'AUTO CONSULTANT');
Insert into category (CAT_ID,CAT_TYPE) values (23804,'CONSULTANT FOR AUTO FINANCE');
Insert into category_match (CAT_ID,CAT_TYPE) values (12790,'AUTO CONSULTANTS');
INSERT INTO CATEGORY_match (CAT_ID,CAT_TYPE) VALUES (23803,'AUTO CONSULTANT');
Insert into category_match (CAT_ID,CAT_TYPE) values (23804,'CONSULTANT FOR AUTO FINANCE');
CREATE INDEX "LOOKING4"."MYINDEX" ON "CATEGORY_MATCH"
    "CAT_TYPE"
  INDEXTYPE IS "CTXSYS"."CONTEXT" ;
CREATE INDEX "LOOKING4"."CAT_TYPE_IDX" ON "CATEGORY"
    "CAT_TYPE"
  INDEXTYPE IS "CTXSYS"."CTXCAT" ;
select cat_id,CAT_TYPE,UTL_MATCH.edit_distance_similarity(CAT_TYPE,'AUTO CONSULTANT') from
select * from category where catsearch(cat_type,
'<query>
      <textquery grammar="context">
      <progression>
      <seq>auto consultant</seq>
      <seq>?(auto) and ?(consultant)</seq>
        </progression>
        </textquery>
</query>'
,NULL)>0
)where rownum<5
23803     AUTO CONSULTANT     100
12790     AUTO CONSULTANTS     94
23804     CONSULTANT FOR AUTO FINANCE     26
update category set cat_type='CONSULTANTS AUTO' WHERE CAT_ID=12790
select cat_id,CAT_TYPE,UTL_MATCH.edit_distance_similarity(CAT_TYPE,'AUTO CONSULTANT') from
select * from category where catsearch(cat_type,
'<query>
      <textquery grammar="context">
      <progression>
      <seq>auto consultant</seq>
      <seq>?(auto) and ?(consultant)</seq>
        </progression>
        </textquery>
</query>'
,NULL)>0
)where rownum<5
23803     AUTO CONSULTANT     100
12790     CONSULTANTS AUTO     32
23804     CONSULTANT FOR AUTO FINANCE     26
select score(1),cat_id,cat_type from CATEGORY_MATCH where cat_id in(
select cat_id from category where catsearch(cat_type,
'<query>
      <textquery grammar="context">
      <progression>
      <seq>auto consultant</seq>
      <seq>?(auto) and ?(consultant)</seq>
        </progression>
        </textquery>
</query>'
,NULL)>0) AND
contains(cat_type,'?(auto) and ?(consultant)',1)>0
9     23803     AUTO CONSULTANT
9     12790     AUTO CONSULTANTS
9     23804     CONSULTANT FOR AUTO FINANCEi have been using catsearch to use progressive relaxation
there are many "cat_types" like "cat_id" =23803,12790 ,the order of words in a sentence changes
there are upto 10 words in each row of "cat_types" column
among others i have referred
Achieving functionality of many preferences using one context index
and
Re: Fuzzy search - more accurate score??
there is very less possibility of repetition of words in a row
utl match seems to work perfect only when the order of appearance of words is same
if you can suggest a way to get a very close score for cat_id 23803 and 12790 it would be much appreciated
thanks and regards

Answer 1

select *
    FROM   (SELECT score(1),score(2),score(3),score(4),GREATEST (SCORE(1), SCORE(2) - 1, SCORE(3) - 2, SCORE(4) - 3) g_scores,
                  UTL_MATCH.EDIT_DISTANCE_SIMILARITY (CAT_TYPE,'AUTO CONSULTANT') EDS,
                  CAT_ID, CAT_TYPE
              FROM   category_match
              WHERE  CONTAINS (cat_type, 'solar water heater* 10 * 10', 1) > 0
              OR     CONTAINS (cat_type, 'NEAR ((?solar, ?water ,?heater), 0, TRUE) * 10 * 10', 2) > 0
             OR     CONTAINS (cat_type, 'NEAR ((?solar, ?water ,?heater), 0, FALSE) * 10 * 10', 3) > 0
             or     CONTAINS (CAT_TYPE, '(?solar AND ?water AND ?heater) * 10 * 10', 4) > 0
             order  by g_scores desc, EDS desc)
   WHERE  ROWNUM<100
100     100     100     100     100     23     4                  SOLAR WATER HEATER-ANU
100     100     100     100     100     22     26901          SOLAR WATER HEATER SUDARSHAN SAUR
100     100     100     100     100     21     30                  SOLAR WATER HEATER INDUSTRIAL
100     100     100     100     100     20     17379          SOLAR WATER HEATER DEALERS-TATA
100     100     100     100     100     20     26906          SOLAR WATER HEATER NUETECH
100     100     100     100     100     20     11465          SOLAR WATER HEATER DEALERS-ANU
100     100     100     100     100     20     21                  SOLAR WATER HEATER-ZING TATA BP
100     100     100     100     100     20     11463          SOLAR WATER HEATER MANUFACTURERS-ANU
100     100     100     100     100     19     8                  SOLAR WATER HEATER MANUFACTURERS
100     100     100     100     100     19     23                  SOLAR WATER HEATER EVACUATED TUBE
100     100     100     100     100     19     49                  SOLAR WATER HEATER-HOTMAX NOVA TATA BP
100     100     100     100     100     19     13357          SOLAR WATER HEATER INDUSTRIAL DEALERS
100     100     100     100     100     18     16300          SOLAR WATER HEATER-TECHNOMAX
100     100     100     100     100     18     9                  SOLAR WATER HEATER DEALERS-TATA BP
100     100     100     100     100     18     20                  SOLAR WATER HEATER-ZING
100     100     100     100     100     18     18                  SOLAR WATER HEATER-ORB SOLAR
100     100     100     100     100     18     22552          SOLAR WATER HEATER-KOTAK URJA
100     100     100     100     100     18     26908          SOLAR WATER HEATER SUPREME
100     100     100     100     100     17     26907          SOLAR WATER HEATER TECHNOMAX"
100     100     100     100     100     17     13322          SOLAR WATER HEATER DISTRIBUTORS
100     100     100     100     100     17     22                  SOLAR WATER HEATER-ETC TATA BP
100     100     100     100     100     17     48                  SOLAR WATER HEATER-VAJRA PLUS TATA BP
100     100     100     100     100     17     27084          SOLAR WATER HEATER SALES
100     100     100     100     100     16     16236          SOLAR WATER HEATER DEALERS-RACOLD
100     100     100     100     100     16     15                  SOLAR WATER HEATER-NUTECH
100     100     100     100     100     16     1                  SOLAR WATER HEATER DEALERS
100     100     100     100     100     15     2                  SOLAR WATER HEATER DEALERS-TATA BP SOLAR
100     100     100     100     100     15     31                  SOLAR WATER HEATER DOMESTIC
100     100     100     100     100     15     13                  SOLAR WATER HEATER DEALERS-V GUARD
100     100     100     100     100     14     17                  SOLAR WATER HEATER-KAMAL SOLAR
100     100     100     100     100     13     11467          SOLAR WATER HEATER DEALERS-GILMA
100     100     100     100     100     13     19                  SOLAR WATER HEATER-GILMA
100     100     100     100     100     13     10                  SOLAR WATER HEATER REPAIRS & SERVICES-TATA SOLAR
100     100     100     100     100     12     10578          SOLAR WATER HEATER
100     100     100     100     100     11     3                  SOLAR WATER HEATER REPAIRS & SERVICES
0      0     100     100     98     25     10120          WATER HEATER SOLAR INDUSTRIAL
0      0     100     100     98     20     12953          WATER HEATER SOLAR-RACOLD
0      0     100     100     98     17     10119          WATER HEATER SOLAR RESIDENCIAL
{code}
the query is working accurately technically
but is there any way to get 10578 on top
the requirement is
---first
solar water heater
solar water heater dealers
solar water heater manufacturers
solar water heater distributors
solar water heater sales
solar water heater repairs and servicing
---followed by
SOLAR WATER HEATER REPAIRS & SERVICES-TATA SOLAR
SOLAR WATER HEATER-KAMAL SOLAR
SOLAR WATER HEATER DEALERS-TATA BP SOLAR   etc etc
so if the end user types in "solar water" the top row would have a row from the table that has what the end user has entered followed by "dealers" or "manufacturers" or "distributors" or "sales" or "repairs and servicing"
so if a row contains "solar water dealer" it shows up on top
or(if "solar water dealer" is not there and "solar water manufacturers" or  "solar water distributors" etc is not present)
a row from the table that has what the end user has entered PLUS "heater" followed by "dealers" or "manufacturers" or "distributors" or "sales" or "repairs and servicing"
so "solar water heater dealers" shows up on top
these words - "dealers" , "manufacturers" , "distributors" , "sales" , "repairs and servicing"  etc remain constant
what i am using right now is
{code}
create or replace
procedure HOME_OLD
p_cat_type in varchar2,
P_LOC IN NUMBER,
P_MAX IN NUMBER,
P_MIN IN NUMBER,
P_OUT OUT SYS_REFCURSOR
as
VARIAB varchar2(500);
VARIAB2 varchar2(500);
VARIAB3 varchar2(500);
VARIAB4 varchar2(500);
begin
--VARIAB2:='?'||replace(P_CAT_TYPE,' ',', ?');
--VARIAB3:='?'||replace(P_CAT_TYPE,' ',' ?');
--DBMS_OUTPUT.PUT_LINE(VARIAB2);
--DBMS_OUTPUT.PUT_LINE(VARIAB3);
SELECT stragg(cat_id) into variab
   FROM   (SELECT GREATEST (SCORE(1), SCORE(2) - 1, SCORE(3) - 2, SCORE(4) - 3) score,
                                  CAT_ID, CAT_TYPE
              FROM   category_match
              -- exact words in order:
              WHERE  CONTAINS (cat_type,get_basic(P_CAT_TYPE), 1) > 0
              -- similar words next to each other in order:
              OR     CONTAINS (cat_type, get_near_syntax(P_CAT_TYPE), 2) > 0
             -- similar words next to each other in any order:
             OR     CONTAINS (cat_type, get_near_syntax_desc(P_CAT_TYPE), 3) > 0
             -- similar words anywhere in any order:
             OR     CONTAINS (cat_type, get_anywhere(P_CAT_TYPE), 4) > 0
             order  by score desc)
   where  rownum < 3;
DBMS_OUTPUT.PUT_LINE(VARIAB);
open p_out
   FOR select * from(select rownum r,name,address1,telephone,mobile,CAT_TYP,cat_id,
(case when  address2=p_loc and ACT_STATUS='Y'  then '1' when  address2=p_loc  then '2' when address2 in
(select NEARBY_LOC from NEAR_BY where LOCALITY_ID=p_loc) and ACT_STATUS='Y' 
then '3' when ADDRESS2 in (select NEARBY_LOC from NEAR_BY where LOCALITY_ID=p_loc)
then '4' when ACT_STATUS='Y' and address2<> p_loc then '5' else '6' end) as marker
  FROM TEST_TEST
  WHERE
  CAT_ID in(select * from table(STRING_TO_TABLE_NUM(variab))) and rownum<P_MAX order by marker) where r>P_MIN;
IF VARIAB IS NULL THEN
OPEN P_OUT
   FOR SELECT * FROM(SELECT rownum r,name,address1,telephone,mobile,CATS
   FROM   (SELECT GREATEST (SCORE(1), SCORE(2) - 1, SCORE(3) - 2, SCORE(4) - 3) score,
                                  NAME,ADDRESS1,TELEPHONE,MOBILE,CATS
              FROM   TEST_TEST2
              -- exact words in order:
              WHERE  CONTAINS (NAME,get_basic(P_CAT_TYPE), 1) > 0
              -- similar words next to each other in order:
              OR     CONTAINS (NAME, get_near_syntax(P_CAT_TYPE), 2) > 0
             -- similar words next to each other in any order:
             OR     CONTAINS (NAME, get_near_syntax_desc(P_CAT_TYPE), 3) > 0
             -- similar words anywhere in any order:
             OR     CONTAINS (NAME, get_anywhere(P_CAT_TYPE), 4) > 0
             ORDER  BY SCORE DESC)
   WHERE  ROWNUM < P_MAX)where r>P_MIN;
END IF;
end home_old;
{code}
the flow is to find what the end user has entered in category table ,if a match exists,find all reg_ids from test_test materialized view that have selected the matched cat_id..
the test_test materialized view lists each company cat_id-selected-by-that-company number of times 
if no match is found in category table what the end user has entered could be a company so  a search in name column of test_test2 materialized view..
this materialized view has one entry for each  company
{code}
create or replace
FUNCTION GET_BASIC(P_CAT_TYPE VARCHAR2)
    RETURN VARCHAR2
  is
  VARIAB2 VARCHAR2(3000);
      begin
VARIAB2:='{'||P_CAT_TYPE||'}*10*10'; 
return(VARIAB2);
END;
create or replace
FUNCTION GET_NEAR_SYNTAX(P_CAT_TYPE VARCHAR2)
    RETURN VARCHAR2
  is
  VARIAB2 VARCHAR2(3000);
      begin
VARIAB2:='NEAR((?{'||replace(P_CAT_TYPE,' ','}, ?{')||'}),10,TRUE)*10*10'; 
return(VARIAB2);
END;
create or replace
FUNCTION GET_NEAR_SYNTAX_DESC(P_CAT_TYPE VARCHAR2)
    RETURN VARCHAR2
  is
  VARIAB2 VARCHAR2(3000);
      begin
VARIAB2:='NEAR((?{'||replace(P_CAT_TYPE,' ','}, ?{')||'}),10,FALSE)*10*10'; 
return(VARIAB2);
END;
{code}
can anything be done to ameliorate this whole flow
can anything be done to eliminate the near_by and act_status and locality checking in ordering by "marker" clause
below is the materialized view creation ddl
SELECT IN_V.REG_ID,
    IN_V.NAME,
    IN_V.TELEPHONE,
    IN_V.MOBILE,
    IN_V.ADDRESS1,
    IN_V.ADDRESS2,
    IN_V.ACT_STATUS,
    resec.cat_id,
    UPPER(STRAGG(IN_V.CAT_TYPE)) AS cat_typ
  FROM
    (SELECT RSC.REG_ID,
      R.NAME,
      RSC.CAT_ID,
      C.CAT_TYPE,
      R.ADDRESS1,
      R.ADDRESS2,
      R.ACT_STATUS,
      R.TELEPHONE,
      R.MOBILE,
      ROW_NUMBER() OVER (PARTITION BY RSC.REG_ID ORDER BY rsc.reg_id) AS TT
    FROM REG_SEG_CAT RSC,
      category C,
      REGISTRATION R
    WHERE C.CAT_ID=RSC.CAT_ID
    AND R.REG_ID  =RSC.REG_ID
    ) IN_V,
    REG_SEG_CAT RESEC
  WHERE in_v.reg_id=resec.reg_id
  AND IN_V.TT      <6
  GROUP BY IN_V.REG_ID,
    IN_V.NAME,
    IN_V.TELEPHONE,
    IN_V.MOBILE,
    IN_V.ADDRESS2,
    IN_V.ACT_STATUS,
    IN_V.ADDRESS1,
    resec.cat_id;
  and sql>desc test_test
REG_ID
NAME
TELEPHONE
MOBILE
ADDRESS1
ADDRESS2
ACT_STATUS
CAT_ID
CAT_TYP
please let me know if you need more info
Edited by: 946207 on Apr 19, 2013 6:22 PM