Word scoring for fuzzy searches
If you have created an index with the WORD_SUPPORT
= PATTERN
index parameter, the search engine considers words
that match your clue approximately and that match your clue exactly.
The search engine uses fuzzy logic to determine whether a pattern
match is considered a hit. It assigns a word score to candidate matches
based on its internal rules.
By default, only words that match your search clue by a relative measure of 70 out of 100—that have a word score of 70 or better—are considered hits. You can change the default by specifying your own value for the WORD_SCORE tuning parameter.
For example, suppose you want the text search engine to count only
as hits words that match your clues by a measure of 85 or better.
You would specify this condition by setting
WORD_SCORE = 85
in
the etx_contains() operator, as shown in the following
example: SELECT id, description FROM videos
WHERE etx_contains ( description,
Row('multimedia document editor' ,
'PATTERN_TRANS & PATTERN_SUBS & WORD_SCORE = 85'));
The following sections show how you perform the types of fuzzy searches enabled by the module.