Oracle Text progressive relaxation

Question

hello,
We're in the process of evaluating Oracle Text search engine so far so good until yesterday when we added Synonyms to our progressive search criterion and it stop working depending on where we place the synonym search. If we place it first everything else stops working (stemming, fuzzy...) If we place it last then the synonym search stops working. I saw a reference to a bug in this conference that seemed similar to the problem, I believe it mentioned that it had been fixed in 10.2.0.3 (this is the version were on).
The following is a sample of plsql code were executing
select score(1), nm_resource, ADDR_RSRC_ST_LN_1, id_resource, ADDR_RSRC_CITY FROM caps_resource where
CONTAINS (nm_resource,
     '<query>
     <textquery lang="ENGLISH" grammar="CONTEXT">' || res_name ||
     '<progression>
<seq><rewrite>transform((TOKENS, ?{?, ?}?, ?AND?))</rewrite></seq>
<seq><rewrite>transform((TOKENS, ??{?, ?}?, ?AND?))</rewrite>/seq>
<seq><rewrite>transform((TOKENS, ?{?, ?}?, ?OR?))</rewrite></seq>
<seq><rewrite>transform((TOKENS, ??{?, ?}?, ?OR?))</rewrite>/seq>
<seq><rewrite>transform((TOKENS, ?{?, ?}?, ?ACCUM?))</rewrite></seq>
<seq><rewrite>transform((TOKENS, ?{?, ?}?, ?NEAR?))</rewrite></seq>
<seq>' || 'SYN(' || REPLACE('' || res_name || '', ' ', ',IMPACT_tst) AND SYN(') || ',IMPACT_tst)' || '</seq>
</progression>
</textquery>
<score datatype="INTEGER" algorithm="default"/>
</query>', 1)>0

Answer 1

Here is a suggested alternative. I have used a syns function that I wrote for another user on another thread, that checks for all combinations of words that could amount to a synonym, up to the maximum number of words that you can specify as an input parameter. I have then used that in a separate contains clause and combined the score results, using the score derived from the synonym only when there is no score from the progressive rewrites, ordering by the progressive rewrites first. Also notice that the ordering must be done in an inner sub-query then the rows limited in an outer subquery. Otherwise it can select the first 100 rows, then order them, instead of the other way around, which can produce an entirely different result set.
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 ('Delagarza,Lorenzo')
  3  INTO caps_resource VALUES ('Diana De La Garza')
  4  INTO caps_resource VALUES ('De La Garza,Fred')
  5  INTO caps_resource VALUES ('somebody else')
  6  SELECT * FROM DUAL
  7  /
4 rows created.
SCOTT@orcl_11g> CREATE INDEX your_index ON caps_resource (nm_resource)
  2  INDEXTYPE IS CTXSYS.CONTEXT
  3  /
Index created.
SCOTT@orcl_11g> BEGIN
  2    CTX_THES.CREATE_THESAURUS ('impact_tst');
  3    CTX_THES.CREATE_RELATION ('impact_tst', 'Delagarza', 'SYN', 'De La Garza');
  4  END;
  5  /
PL/SQL procedure successfully completed.
SCOTT@orcl_11g> create or replace function syns
  2    (p_words in varchar2,
  3       p_thes     in varchar2,
  4       p_num     in number default 3) -- maximum number of words per synonym phrase
  5       return varchar2
  6  as
  7    v_words_in      varchar2 (32767) := ltrim (p_words) || ' ';
  8    v_words_out     varchar2 (32767);
  9  begin
10    while instr (v_words_in, '  ') > 0 loop
11        v_words_in := replace (v_words_in, '  ', ' ');
12    end loop;
13    while length (v_words_in) > 1
14    loop
15        for i in reverse 1 .. least (p_num, (length (v_words_in) - length (replace (v_words_in, ' ', ''))))
16        loop
17          if instr (ctx_thes.syn (substr (v_words_in, 1, instr (v_words_in, ' ', 1, i) - 1), p_thes), '|') > 0
18            or i = 1 then
19            v_words_out := v_words_out
20            || ' AND ('
21            || ctx_thes.syn (substr (v_words_in, 1, instr (v_words_in, ' ', 1, i) - 1), p_thes)
22            || ')';
23            v_words_in := substr (v_words_in, instr (v_words_in, ' ', 1, i) + 1);
24            exit;
25          end if;
26        end loop;
27    end loop;
28    return ltrim (v_words_out, ' AND ');
29  end syns;
30  /
Function created.
SCOTT@orcl_11g> show errors
No errors.
SCOTT@orcl_11g> VARIABLE res_name VARCHAR2 (100)
SCOTT@orcl_11g> EXEC :res_name := 'De La Garza'
PL/SQL procedure successfully completed.
SCOTT@orcl_11g> SELECT syns (:res_name, 'impact_tst') FROM DUAL
  2  /
SYNS(:RES_NAME,'IMPACT_TST')
({DE LA GARZA}|{DELAGARZA})
SCOTT@orcl_11g> SELECT the_score, nm_resource
  2  FROM   (select DECODE (score(1), 0, SCORE(2), SCORE(1)) AS the_score, nm_resource
  3            FROM   caps_resource
  4            where  CONTAINS (nm_resource,
  5            '<query>
  6            <textquery lang="ENGLISH" grammar="CONTEXT">' || :res_name ||
  7             '<progression>
  8                <seq><rewrite>transform((TOKENS, "{",  "}", "AND"))</rewrite></seq>
  9                <seq><rewrite>transform((TOKENS, "?{", "}", "AND"))</rewrite>/seq>
10                <seq><rewrite>transform((TOKENS, "{",  "}", "OR"))</rewrite></seq>
11                <seq><rewrite>transform((TOKENS, "?{", "}", "OR"))</rewrite>/seq>
12                <seq><rewrite>transform((TOKENS, "{",  "}", "ACCUM"))</rewrite></seq>
13                <seq><rewrite>transform((TOKENS, "{",  "}", "NEAR"))</rewrite></seq>
14              </progression>
15            </textquery>
16            <score datatype="INTEGER" algorithm="default"/>
17             </query>', 1) > 0
18            OR     CONTAINS (nm_resource, syns (:res_name, 'impact_tst'), 2) > 0
19            ORDER  BY SCORE (1) DESC, SCORE (2) DESC)
20  WHERE  ROWNUM < 100
21  /
THE_SCORE NM_RESOURCE
        76 Diana De La Garza
        76 De La Garza,Fred
         5 Delagarza,Lorenzo
SCOTT@orcl_11g>