As with a surprising number of similar issues relating to human language, there's already a module on CPAN for this:
Lingua::EN::Inflect -- behold:
#!/usr/bin/env perl
use strict;
use warnings;
use Lingua::EN::Inflect 'A';
while (<DATA>) {
while ( /\b (an? \s (\w+))/gx ) {
my ( $phrase, $word ) = ( $1, $2 );
if ( $phrase eq A( $word )) {
print "RIGHT: $phrase\n";
}
else {
print "WRONG: $phrase\n";
}
}
}
__DATA__
I am a howling maniac in an hospital.
It takes a hour to win an honorable mention.
Calling him a umpire is a euphemism.
This is a united front against a evil empire.
That just shows how to use the module to check for errors. It should be easy to tweak the example a bit so as to fix them.
(updated to fix a typo in the cpan link.)
One more update: a bit more logic is needed to handle various kinds and amounts of punctuation that might occur between the article and the word that follows:
while (<DATA>) {
while ( /\b ((an?) \s\W* (\w+))/gx ) {
my ( $phrase, $article, $word ) = ( $1, $2, $3 );
my $bigram = "$article $word";
if ( $bigram eq A( $word )) {
print "RIGHT: $phrase\n";
}
else {
print "WRONG: $phrase\n";
}
}
}
__DATA__
Suitable for a (older) child or an 'early adult'.