It's not clear to me how, if you want to maintain the word lists strictly in parallel, your desired output isn't just @words1.

I think you might want something like this:

my %xlate;
# open DOC1 here
while (<DOC1>) {
  chomp;
  my($word_part) = split /_/;
  $xlate{$word_part} = $_;
}
# open DOC2 here
while (<DOC2>) {
  s/(\S+)/$xlate{$1}/ge;
  print;
}
[download]

Though that won't strip any newlines. Doing so isn't difficult.