Let me describe a quick way of doing this...

0. assumption - you will never modify file 1, because you will only delete from file 2;

1. read the first file into a hash table, having each word as the hash key;

2. create a third file;

3. while scanning the second file, check the hash table built in step 1 for existance of the word;
if the word exists, do not print to the third file;
if the word does not exist, print the the third file;

4. replace file 2 with the third file.

#!/usr/bin/perl -w
use strict;
use IO::File;

my %hash = ();

my $f = IO::File->new("file1.txt", "r") or die "can not open file 1";
while (my $line = <$f>) {
  chomp $line;
  for my $word (split /\s*,\s*/, $line) {
     $hash{$word}++;
  }
}

my $f3 = IO::File->new("file3.txt", "w") or die "can not create file 3
+";

my $f2 = IO::File->new("file2.txt", "r") or die "can not open file 2";

while (my $line = <$f2>) {
  chomp $line;
  my @words = ();
  for my $word (split /\s*,\s*/, $line) {
    if (! exists $hash{$word}) {
      push @words, $word;
  }
  print $f3 join(",", @words), "\n";
}

undef $f;
undef $f2;
undef $f3;

# then replace file 2 with file 3...
[download]