The issue arises from what constitutes a "word". The basic definition is @words = split ' ', $line;, which works in most cases. However, it's arguable that a better definition could be @words = $line =~ /(\b\w+(?:['-]\w+)\b)/;. Of course, you're now depending on the definition of \w, which includes underscore and doesn't include apostrophe or hyphen. *shrugs* YMMV a huge amount. Parsing any natural language is much harder than parsing Perl which, as everyone knows, can't be done in Perl.

Good luck! I mean it.