how to extract unique words from cell and count them - sql

Question

I have an column "DESCRIPTION" (VARCHAR2 (500 Byte))
I want as result two columns. First extract from each cell unique words and display them in one column, and in second count their frequency.
Additionaly I have limiting parametre "ENTRYDATE" (i.e. "WHERE ENTRYDATE BETWEEN 20180101 and 20190101"). Because table is quite big.
I have some solution in Excel, but it's messy and painful to do.
Is it even possible to do in Oracle with SELECT?
Example:
NUMBER OF COLUMN | EXPLANATION
1 | roses are red violets are blue
2 | red violets
3 | red
4 | roses
5 | blue
RESULT:
WORDS | COUNTING
roses | 2
are | 2
red | 3
violets | 2
blue | 2
Variation of query:
with test as
(select 1 as nor, 'roses are red violets are blue' as explanation from dual union all
select 2 as nor, 'red violets' as explanation from dual union all
select 3 as nor, 'red' as explanation from dual union all
select 4 as nor, 'roses' as explanation from dual union all
select 5 as nor, 'blue' as explanation from dual
),
temp as
(select nor,
trim(column_value) word
from test join xmltable(('"' || replace(explanation, ' ', '","') ||'"')) on 1 = 1
)
select word,
count(*)
from temp
group by word
order by word;
returns ORA-00905: missing keyword

Answer 1

Split explanation into rows (so that you'd get words), then apply COUNT function to those words.
SQL> with test (nor, explanation) as
2 (select 1, 'roses are red violets are blue' from dual union all
3 select 2, 'red violets' from dual union all
4 select 3, 'red' from dual union all
5 select 4, 'roses' from dual union all
6 select 5, 'blue' from dual
7 ),
8 temp as
9 (select nor,
10 regexp_substr(explanation, '[^ ]+', 1, column_value) word
11 from test join table(cast(multiset(select level from dual
12 connect by level <= regexp_count(explanation, ' ') + 1
13 ) as sys.odcinumberlist)) on 1 = 1
14 )
15 select word,
16 count(*)
17 from temp
18 group by word
19 order by word;
WORD COUNT(*)
------------------------------ ----------
are 2
blue 2
red 3
roses 2
violets 2
SQL>
You mentioned entrydate column but there's none in your sample data so - if necessary, include it into the TEMP CTE.
Edit
Huh, Oracle 9i ... back to the Dark Ages:
SQL> with test (nor, explanation) as
2 (select 1, 'roses are red violets are blue' from dual union all
3 select 2, 'red violets' from dual union all
4 select 3, 'red' from dual union all
5 select 4, 'roses' from dual union all
6 select 5, 'blue' from dual
7 ),
8 temp as
9 (select nor,
10 trim(column_value) word
11 from test join xmltable(('"' || replace(explanation, ' ', '","') ||'"')) on 1 = 1
12 )
13 select word,
14 count(*)
15 from temp
16 group by word
17 order by word;
WORD COUNT(*)
-------------------- ----------
are 2
blue 2
red 3
roses 2
violets 2
SQL>

Answer 2

The problem is in your old Oracle version. This query should work, it has only basic connect by, instr and dbms_random:
select word, count(1) counting
from (
select id, trim(case pos2 when 0 then substr(description, pos1)
else substr(description, pos1, pos2 - pos1)
end) word
from (
select id, description,
case level when 1 then 1 else instr(description, ' ', 1, level - 1) end pos1,
instr(description, ' ', 1, level) pos2
from t
connect by prior dbms_random.value is not null
and prior id = id
and level <= length(description) - length(replace(description, ' ', '')) + 1))
group by word
demo

Answer 3

-- Oracle 12c+
with test (nor, explanation) as (
select 1, 'roses are red violets are blue' from dual union all
select 2, 'red violets' from dual union all
select 3, 'red' from dual union all
select 4, 'roses' from dual union all
select 5, 'blue' from dual)
select regexp_substr(explanation, '\S+', 1, lvl) word, count(*) cnt
from test,
lateral(
select rownum lvl
from dual
connect by level <= regexp_count(explanation, '\S+')
)
group by regexp_substr(explanation, '\S+', 1, lvl);
WORD CNT
------------------------------ ----------
roses 2
are 2
violets 2
red 3
blue 2