Thanks for picking up the error - I had actually found and fixed the bug in my testing but apparently pasted in the broken version. Doh!

As to why Damian chose to include a list of the full words and then incur the processing overhead of chopping them down, my best guess is to make it easier to maintain the list of words. A more efficient solution would be to generate the Lingua::EN::Inflect module (using the form you suggest) from the list of full words at the time the module was installed.