Limiting to specific length from a particular text

Lets say I have a field of long text.

Every time the word "sentence" shows, we will trim the data from the word "sentence" to 50 characters.

Output:

I tried combination of substring and instring, but didn't work out

If you don't have/want/know regexp you could use significantly faster old functions SubStr() And InStr() to get your expected result.
With sample data as:

WITH      --  S a m p l e    D a t a :
	tbl (COL_A, COL_B) AS
		(	Select  'S101', 
              '	Sentence 1: There once was a poor boy who spent his days going door-to-door selling newspapers to pay for school. 
              One day, as he was walking his route, he started feeling low and weak.The poor boy was starving, so he decided 
              to ask for food when he came to the next door. 
              Sentence 2: The poor boy asked for food but was denied every time, 
              until he reached the door of a girl. He asked for a glass of water,but seeing his poor state, the girl came back 
              with a glass of milk. The boy asked how much he owed her for the milk, but she refused payment.
              Sentence 3: Years later, the girl, who was now a grown woman, fell sick. She went from doctor to doctor, but no one 
              was able to cure her.Finally, she went to the best doctor in town.
              Sentence 4: The doctor spent months treating her until she was finally cured. Despite her happiness, she was afraid 
              she couldn’t afford to pay the bill. But, when the hospital handed her the bill, it read, ‘Paid in full, with a glass of milk.’'
      From Dual 
		)

Subquery limiting LEVEL connections calculates number of occurencies of string 'Sentence ' in COL_B. Used in code with InStr() function.
Option 1: Return each shortened sntence in it's own row:

Select  COL_A,
        SubStr(COL_B, InStr(COL_B, 'Sentence ', 1, LEVEL), 50) "COL_B"
From    tbl
Connect By LEVEL <= ( Select (Length(COL_B) - Length(REPLACE(COL_B, 'Sentence ')) ) / Length('Sentence ') "SENTENCES" From tbl )

--  COL_A  COL_B                                             
--  -----  --------------------------------------------------
--  S101   Sentence 1: There once was a poor boy who spent hi
--  S101   Sentence 2: The poor boy asked for food but was de
--  S101   Sentence 3: Years later, the girl, who was now a g
--  S101   Sentence 4: The doctor spent months treating her u

Option 2: Return all shortened sentences in 1 row (be aware of 4000 chars limit for VarChar2 datatype):

Select  COL_A,
        LISTAGG(SubStr(COL_B, InStr(COL_B, 'Sentence ', 1, LEVEL), 50), Chr(10)) WITHIN GROUP (ORDER BY LEVEL) OVER(Partition By COL_A) "COL_B"
From    tbl
Connect By LEVEL <= ( Select (Length(COL_B) - Length(REPLACE(COL_B, 'Sentence ')) ) / Length('Sentence ') "SENTENCES" From tbl )
FETCH FIRST ROW ONLY
--  
--  R e s u l t:
--  COL_A  COL_B
--  -----  --------------------------------------------------
--  S101   Sentence 1: There once was a poor boy who spent hi 
--         Sentence 2: The poor boy asked for food but was de 
--         Sentence 3: Years later, the girl, who was now a g 
--         Sentence 4: The doctor spent months treating her u

Note: Used Chr(10) as separator to get result in lines for better readability, you could use anything else that suits you.
ADDITION:
For multi rows you should filter the subquery in Connect By clause by key column of the table and either use DISTINCT keyword (could be performance costly with large number of rows) or do the aggregation on COL_B (Max or Min) grouped by COL_A. Both (Distinct and Aggregation) will remove duplicate rows (you don't need Fetch First in Option 2). Here is the code:

-- Option 1
Select  DISTINCT t.COL_A,
        SubStr(t.COL_B, InStr(t.COL_B, 'Sentence ', 1, LEVEL), 50) "COL_B"
From    tbl t
Connect By LEVEL <= ( Select (Length(COL_B) - Length(REPLACE(COL_B, 'Sentence ')) ) / Length('Sentence ') "SENTENCES" From tbl Where COL_A = t.COL_A)        
Order By COL_A, COL_B        
      
-- Option 2        
Select  DISTINCT t.COL_A,
        LISTAGG(SubStr(t.COL_B, InStr(t.COL_B, 'Sentence ', 1, LEVEL), 50), Chr(10)) WITHIN GROUP (ORDER BY LEVEL) OVER(Partition By t.COL_A) "COL_B"
From    tbl t
Connect By LEVEL <= ( Select (Length(COL_B) - Length(REPLACE(COL_B, 'Sentence ')) ) / Length('Sentence ') "SENTENCES" From tbl Where COL_A = t.COL_A)
Order By COL_A

One option might be to split that long text into rows (one row per sentence), extract 50 characters from each row and aggregate them back.

Sample data:

SQL> with test (col) as
  2  (select 'Sentence 1: There once was a poor boy who spent his days going door-to-door selling
  3  newspapers to pay for school. One day, as he was walking his route, he started feeling low and weak.
  4  The poor boy was starving, so he decided to ask for food when he came to the next door.
  5  Sentence 2: The poor boy asked for food but was denied every time, until he reached the door of a girl.
  6  He asked for a glass of water,but seeing his poor state, the girl came back with a glass of milk.
  7  The boy asked how much he owed her for the milk, but she refused payment.
  8  Sentence 3: Years later, the girl, who was now a grown woman, fell sick. She went from doctor
  9  to doctor, but no one was able to cure her.Finally, she went to the best doctor in town.
 10  Sentence 4: The doctor spent months treating her until she was finally cured. Despite her
 11  happiness, she was afraid she couldn’t afford to pay the bill. But, when the hospital
 12  handed her the bill, it read, ‘Paid in full, with a glass of milk.' from dual),

Query begins here:

 13  temp as
 14    (select level lvl,
 15            regexp_substr(replace(col, 'Sentence', '#'), '[^#]+', 1, level) val
 16     From test
 17     connect by level <= regexp_count(col, 'Sentence')
 18    )
 19  select listagg(substr('Sentence ' || val, 1, 51), ' ') within group (order by lvl) result
 20  from temp;

RESULT
--------------------------------------------------------------------------------
Sentence  1: There once was a poor boy who spent hi Sentence  2: The poor boy as
ked for food but was de Sentence  3: Years later, the girl, who was now a g Sent
ence  4: The doctor spent months treating her u


SQL>

I thought using a regexp without lookahead or lookbehind (as it is not supported):

select
  REGEXP_REPLACE(
    'Sentence 1: There once was a poor boy who spent his days going door-to-door selling newspapers to pay for school. One day, as he was walking his route, he started feeling low and weak.The poor boy was starving, so he decided to ask for food when he came to the next door. Sentence 2: The poor boy asked for food but was denied every time, until he reached the door of a girl. He asked for a glass of water,but seeing his poor state, the girl came back with a glass of milk. The boy asked how much he owed her for the milk, but she refused payment.Sentence 3: Years later, the girl, who was now a grown woman, fell sick. She went from doctor to doctor, but no one was able to cure her.Finally, she went to the best doctor in town.Sentence 4: The doctor spent months treating her until she was finally cured. Despite her happiness, she was afraid she couldn’t afford to pay the bill. But, when the hospital handed her the bill, it read, ‘Paid in full, with a glass of milk.’',
    '(: )(.{1,38})(.*?)(Sentence\s+\d+|$)',
    '\1\2 \4'
  ) from dual

The idea was that we know that it starts with Sentence followed by a number
and then : so we could assume that we first start by searching : and then capture what is before the next Sentense # (where # is some digits).

We capture the text in the second group with a security in case it is shorter than expected. Then we capture the text to remove and then match the next sentence delimiter.

You can test it here: https://dbfiddle.uk/kGQMyNx_

@Littlefoot's answer mostly works but relies on replacing 'Sentence' with '#' to make the regex '[^#]+' work, which means it would fail when the input contains '#' itself.

A more robust approach would be to use the instr function instead to find the positions of occurrences of 'Sentence' in the input:

with sentences as (
    select
    	'Sentence 1: There once was a poor boy who spent his days going door-to-door selling newspapers to pay for school. One day, as he was walking his route, he started feeling low and weak.The poor boy was starving, so he decided to ask for food when he came to the next door. Sentence 2: The poor boy asked for food but was denied every time, until he reached the door of a girl. He asked for a glass of water,but seeing his poor state, the girl came back with a glass of milk. The boy asked how much he owed her for the milk, but she refused payment.Sentence 3: Years later, the girl, who was now a grown woman, fell sick. She went from doctor to doctor, but no one was able to cure her.Finally, she went to the best doctor in town.Sentence 4: The doctor spent months treating her until she was finally cured. Despite her happiness, she was afraid she couldn’t afford to pay the bill. But, when the hospital handed her the bill, it read, ‘Paid in full, with a glass of milk.’'
    	as s from dual
)
select listagg(sentence, ' ') within group (order by lvl)
from (
    select substr(s, instr(s, 'Sentence', 1, level), 50) sentence, level lvl
    from sentences
	connect by level <= regexp_count(s, 'Sentence')
);

This outputs:

Sentence 1: There once was a poor boy who spent hi Sentence 2: The poor boy asked for food but was de Sentence 3: Years later, the girl, who was now a g Sentence 4: The doctor spent months treating her u

Demo: https://livesql.oracle.com/apex/livesql/s/o8l4vu3rzs67vs8ag2ldmzs6k

You can use a recursive query to find each sub-string starting with Sentence \d+: and recursively concatenate up to 50 characters of those sub-strings:

WITH data (rn, column_a, column_b, epos, updated_column_b) AS (
  SELECT ROWNUM,
         column_a,
         column_b,
         1,
         EMPTY_CLOB()
  FROM   table_name
UNION ALL
  SELECT rn,
         column_a,
         column_b,
         REGEXP_INSTR(column_b, 'Sentence \d+:', epos, 2),
         updated_column_b
         || ' '
         || SUBSTR(
              column_b,
              epos,
              CASE REGEXP_INSTR(column_b, 'Sentence \d+:', epos, 2)
              WHEN 0
              THEN 50
              ELSE LEAST(
                     50,
                     REGEXP_INSTR(column_b, 'Sentence \d+:', epos, 2) - epos
                   )
              END
            )
  FROM   data
  WHERE  epos > 0
)
SEARCH DEPTH FIRST BY rn SET order_id
SELECT column_a,
       SUBSTR(updated_column_b, 2) AS updated_column_b
FROM   data
WHERE  epos = 0;

Which, for the sample data:

CREATE TABLE table_name (Column_A, Column_B) AS
SELECT 'S101',
       EMPTY_CLOB() || 'Sentence 1: There once was a poor boy who spent his days going door-to-door selling newspapers to pay for school. One day, as he was walking his route, he started feeling low and weak.The poor boy was starving, so he decided to ask for food when he came to the next door. Sentence 2: The poor boy asked for food but was denied every time, until he reached the door of a girl. He asked for a glass of water,but seeing his poor state, the girl came back with a glass of milk. The boy asked how much he owed her for the milk, but she refused payment.Sentence 3: Years later, the girl, who was now a grown woman, fell sick. She went from doctor to doctor, but no one was able to cure her.Finally, she went to the best doctor in town.Sentence 4: The doctor spent months treating her until she was finally cured. Despite her happiness, she was afraid she couldn’t afford to pay the bill. But, when the hospital handed her the bill, it read, ‘Paid in full, with a glass of milk.’'
FROM   DUAL

Which outputs:

fiddle

Here is a regex that catches every "sentence up to 50 characters":

Sentence.{50}

I hope it will help.