Question

If I have a table like this…

create table #words (
id int identity,
word varchar(1024)
)

insert into #words (word) values ('dock')
insert into #words (word) values ('dockable')

and i do a LIKE query

select id, word from #words where word like '%dock%'

Is there a way to tell which result would be the most accurate?

Answer 1

For complex multi-word criteria you should use Full Text Search and CONTAINSTABLE. The output of this table function contains a RANK column:

The table produced by CONTAINSTABLE
includes a column named RANK. The RANK
column is a value (from 0 through
1000) for each row indicating how well
a row matched the selection criteria.
This rank value is typically used in
one of these ways in the SELECT
statement:

• In the ORDER BY clause to return the highest-ranking rows as the first rows
  in the table.
• In the select list to see the rank value assigned to each row.

For simple single word criteria you should implement a Levenshtein distance function in SQL CLR and use that to find the most similar best match words (or use the one from Ken Redler’s linked project).