If I understand well you don't simply need to replace ',' with a space, but also to remove duplicates in a smarter way.
If I modify that expression to work with space instead of ',', I get
select regexp_replace('A B A A C D' ,'([^ ]+)( [ ]*\1)+', '\1') from dual
which gives 'A B A C D', not what you need.
A way to get your needed result could be the following, a bit more complicated:
with string(s) as ( select 'A B A A C D' from dual)
select listagg(case when rn = 1 then str end, ' ') within group (order by lev)
from (
select str, row_number() over (partition by str order by 1) rn, lev
from (
SELECT trim(regexp_substr(s, '[^ ]+', 1, level)) str,
level as lev
FROM string
CONNECT BY instr(s, ' ', 1, level - 1) > 0
)
)
My main problem here is that I'm not able to build a regexp that checks for non adjacent duplicates, so I need to split the string, check for duplicates and then aggregate again the non duplicated values, keeping the order.
If you don't mind the order of the tokens in the result string, this can be simplified:
with string(s) as ( select 'A B A A C D' from dual)
select listagg(str, ' ') within group (order by 1)
from (
SELECT distinct trim(regexp_substr(s, '[^ ]+', 1, level)) as str
FROM string
CONNECT BY instr(s, ' ', 1, level - 1) > 0
)
