The following query will return Triple word from a table like the
following:

pretty, woman, a
a, pretty woman
china,great, wall
wall, great, china
great, wall, china

=============================
How can I elimate rows that contain the same words?
The output should be:

pretty, woman, a
wall, great, china
-OR-
a, pretty woman
wall, great, china
-OR-
a, pretty woman
great, wall, china

.... etc

=============================
The output should not have 2 row containsexact 3 words

For example, if I have 3 words ( pretty + woman + a )
pretty, woman, a

Then the output will not contain (a + pretty + woman)
a, pretty woman

=============================
SELECT DISTINCT A.word As word1, B.word AS word2, C.word AS word3
FROM Data AS A
JOIN Data AS B
ON A.document = B.document
AND A.position < B.position
AND ABS(A.position - B.position) < 3
JOIN Data AS C
ON A.document = C.document
AND A.position < C.position
AND B.position < C.position
AND ABS(A.position - C.position) < 3
ORDER BY word1;

Carl Forsman wrote:
Just in case you did not see this post in comp.databases.mysql

The question to the OP is really "what problem are you trying to solve?"
To what is the position relative? is there also a DOCID in there
somewhere... having a word and a position, it must be related somehow to
the doc. In this case, again, you have really not provided sufficient
information to actually address the real problem.

Again, because I do not have all of the facts, I must make some educated
guesses based on the OP's request.

Your WORDS table ""appears"" to be:

docid,word,position.

If this is the case you still need to fix your table definition such
that your PK=(docid,word) This way you can NEVER have the same words
twice eliminating your need to come up with some convoluted method to
ensure uniqueness. In other "words" your WORDS table can never have
duplicate "words" for the same id.