The word borders are tricky and certainly should not be used with Unicode data, as /\b/ means a border between \W and \w, i.e. between [A-Za-z0-9] and [^A-Za-z0-9].

You must decide what you call a 'word'. In my opinion it is any alphanumeric sequence with apostrophes and hyphens, but your definition may differ.

#!/usr/bin/perl

use warnings;
use strict;

my $test = "TEST TESt TE'st T 12TE";
while ($test =~ /(?<![\pL\pN\'-])       #NOT a hyphen, apostroph, lett
+er or number before
                 (\p{Lu}{2,})           # two or more uppercase letter
+s
                 (?![\pL\pN\'-])        #NOT a hyphen, apostroph, lett
+er or number after
                /xg) 
{  
    print "$1\n";
}
[download]

My regex matches only "TEST" at the beginning of the string in the example.