Here's a start. It works with your sample input, but it includes no error checking, so it's not very robust. The internal case analysis can probably be tightened.

the lowliest monk

#!/usr/bin/perl -w

use strict;
use Data::Dumper;

my %result;

for my $l (<DATA>) {
  chomp $l;
  my ($cod, $desc) = split ' ', $l, 2;
  my @parts = split //, $cod;
  my $key = '';
  my $hash = \%result;
  while (1) {
    $key .= shift @parts;
    if (@parts) {
      if ( ref $hash->{ $key } ) {
        $hash = $hash->{ $key };
      }
      else {
        $hash = ( $hash->{ $key } = +{ 0 => $hash->{ $key } } );
      }
    }
    else {
      if ( defined $hash->{ $key } ) {
        $hash->{ $key }->{ 0 } = $desc;
      }
      else {
        $hash->{ $key } = $desc;
      }
      last;
    }
  }
}
print Dumper \%result;
__DATA__
1 ITEM
11 Sub Item 1
111 Item X
112 Item
1121 Another Item
11212 And Another Item
12 Sub Item 2
[download]