I want to use Find and Replace to delete page numbers which consists of having 1 to 3 character numbers each in its paragraph.

I also would like to unite the paragraphs above and bellow the page number if the paragraph from bellow starts with a low cap, or point, or ! , ? etc....

Can you post some examples of both cases?

Sorry, I didn't explained it well.
So they are not page numbers( well, they were before I converted the file from .epub to .doc ).

Example of the second case:

After a shocked pause for a couple of years, one of the
12
clump said: ‘He just went! Just like that! One day he was here, next he was gone!’ If the other trees had been humans, they would have shuffled their feet.

What I want is to delete all numbers in such case ^ (numbers that are a paragraph on their own, not within a a paragraph with text) , and unite the upper and lower paragraph IF the lower paragraph starts with low cap.

Try using the following wildcard Find/Replace:
Find:
([!0-9]{1,3})^13[0-9]{1,3}^13
Replace:
\1^32

Now I have this problem :
https://drive.google.com/file/d/0B3W...it?usp=sharing

It also selects a few chars from the paragraph which is above.

Have you tried actually running the replace operation? It should be OK. Word will select the characters preceding the first paragraph end mark, but they will be replaced with themselves, i.e. they won’t be removed (this is done through “\1” in the Replace With box). I successfully tested this wildcard pattern in Word 2007 on the sentence you gave as example.

Note that you can always undo any replace operation through “Ctrl+Z”.

Also note that depending on your system locale and the active list separator, the semi-colons within curly brackets {1;3} might have to be replaced with a simple comma {1,3}.

You could try using this simpler pattern:
Find:
^13[0-9]{1,3}^13
Replace:
^32

Thanks, the first one worked exactly like you said.
I am grateful for your help.