Alternative scoring for CONTAINS operator?

Question

I am trying to figure out how to query using CONTAINS and get slightly different result set scores than the default is returning. By default, it looks like rows are given the same score as long as they contain the same number of the search terms. For example, given the following table:
ID TEXT
1 Eyes Wide Open
2 Open Your Eyes
3 My Eyes Are Wide Open
4 Eyes Open
5 Open Eyes
If I select based on a CONTAINS of "eyes open", all of the rows are returned with the same weight (which I think I understand based on the description of the default scoring algorithm).
But what I would like is a scoring algorithm that takes into account the number of words matched vs. the total number of words in the column, and preferrably also considers the order of the words in the column to the order they were searched on.
So given the previous table, if I select based on a CONTAINS of "eyes open", I would expect the scores to rank ID 4 the highest, followed by 5, 1, 2, 3. (In this case all of the column words are significant and I am using an empty stoplist).
Does anyone know how to accomplish this scoring behavior? Do I need to implement a new scoring algorithm, or can this be accomplished using a syntax for CONTAINS that I just don't know about?
Thanks for your help in advance.
-Chris

Answer 1

I don't see any way to get exactly what you want using Text, but the method below should do it.
SCOTT@10gXE> CREATE TABLE your_table
  2    (id   NUMBER,
  3       text VARCHAR2 (30))
  4  /
Table created.
SCOTT@10gXE> INSERT ALL
  2  INTO your_table VALUES (1, 'Eyes Wide Open')
  3  INTO your_table VALUES (2, 'Open Your Eyes')
  4  INTO your_table VALUES (3, 'My Eyes Are Wide Open')
  5  INTO your_table VALUES (4, 'Eyes Open')
  6  INTO your_table VALUES (5, 'Open Eyes')
  7  SELECT * FROM DUAL
  8  /
5 rows created.
SCOTT@10gXE> CREATE OR REPLACE FUNCTION list_element
  2       (p_string    VARCHAR2,
  3        p_element   INTEGER,
  4        p_separator VARCHAR2 DEFAULT ' ')
  5       RETURN          VARCHAR2
  6  AS
  7    v_string      VARCHAR2 (32767);
  8  BEGIN
  9    v_string := p_string || p_separator;
10    FOR i IN 1 .. p_element - 1 LOOP
11        v_string := SUBSTR (v_string,
12                      INSTR (v_string, p_separator)
13                        + LENGTH (p_separator));
14    END LOOP;
15    RETURN SUBSTR (v_string, 1, INSTR (v_string, p_separator) - 1);
16  END list_element;
17  /
Function created.
SCOTT@10gXE> VARIABLE search_words VARCHAR2 (30)
SCOTT@10gXE> EXEC :search_words := 'eyes open'
PL/SQL procedure successfully completed.
SCOTT@10gXE> SELECT id, text,
  2           ROUND (100 * COUNT (*) / num_words) AS hit_score,
  3           SUM (order_score) AS order_score,
  4           ROUND (100 * COUNT (*) / num_words)
  5             - SUM (order_score) AS combo_score
  6  FROM   (SELECT id, text,
  7                (LENGTH (text) - LENGTH (REPLACE (text, ' ', '')) + 1) AS num_words,
  8                CASE WHEN position > NVL (LAG (position) OVER
  9                                (PARTITION BY id ORDER BY word_order), 0)
10                  THEN 0
11                  ELSE 1
12                END AS order_score
13           FROM   (SELECT id, text, words, word_order,
14                    INSTR (UPPER (text), words) AS position
15                FROM   your_table,
16                    (SELECT list_element (UPPER (:search_words), ROWNUM) AS words,
17                         ROWNUM AS word_order
18                     FROM   DUAL
19                     CONNECT BY LEVEL <= (LENGTH (:search_words)
20                                 - LENGTH (REPLACE (:search_words, ' ', ''))) + 1)
21                WHERE  INSTR (UPPER (text), words) > 0))
22  GROUP BY id, text, num_words
23  ORDER BY hit_score DESC, order_score
24  /
        ID TEXT                            HIT_SCORE ORDER_SCORE COMBO_SCORE
         4 Eyes Open                             100           0         100
         5 Open Eyes                             100           1          99
         1 Eyes Wide Open                         67           0          67
         2 Open Your Eyes                         67           1          66
         3 My Eyes Are Wide Open                  40           0          40
SCOTT@10gXE>