Is there an even cooler way to this?

You mean, is there some other way that is less brittle, ad-hoc and prone to outrageous failure? I should hope so.

But it would be hard to suggest decent improvements, because we don't have any reason to believe that line 4 of  $F[0].xml must always be the place where your new  <keywords> $F[1] </keywords> string should properly replace the original line content.

But a few things that would simplify your approach:

• add "-l" to the perl options (perl -lane), and stop worrying about line-feed characters
• read the whole input file into an array and close it
• replace the fourth array element (index 3) with your new keyword stuff
• open original file name again for output and write the array to it

then hope you didn't just obliterate your xml data, given that you haven't done any error checking, XML parsing, yadda yadda...