A quick and simple way, especially if you don't want to read the whole file into memory first, would be:

s/([A-Z0-9]{5,})/$seenit{$1}++ or print "$1\n"/egi while <>;
[download]

Better yet, save the printing until the end, so you can sort the words alphabetically, or perhaps by the number of appearances:

s/([A-Z0-9]{5,})/$seenit{$1}++/egi while <>;

## Sorted by name
for (sort keys %seenit) {
  print "$_: $seenit{$_}\n";
}

## Sorted by freuency, then by name:
for (sort {$seenit{$a} <=> $seenit{$b} or $a cmp $b} keys %seenit) {
  print "$_: $seenit{$_}\n";
}
[download]

As a final suggestion, you may want to disregard the case of the words, in which case you'd want to use $seenit{lc $1}. Probably best, as words at the start of a sentence tend to be capitalized.