Hello nikolay,

I would store the exceptions in two hashes, then use exists to test whether a given word should be excluded. I begin by splitting the input string on whitespace. Note the presence of a capture group in the regular expression given to split: the captured string (in this case, the whitespace) is added to the list returned by split, to enable the string to be reassembled correctly after the substitutions have been performed.

#! perl
use strict;
use warnings;

my %exclude_left  = map { lc $_ => undef } qw( web 2 );
my %exclude_right = map { lc $_ => undef } qw( program call );

my $z = 'Web-developer, perl-program, explicit-element, function-call,
+ 2-x speed.';
my @phrases = split /(\s+)/, $z;

for (@phrases)
{
    #     1 2       3           <-- capture groups
    if (/ ( (\w+) - (\w+) ) /x)
    {
        my ($phrase, $left, $right) = ($1, $2, $3);

        s/$phrase/$right-$left/ unless exists $exclude_left {lc $left 
+} ||
                                       exists $exclude_right{lc $right
+};
    }
}

print '|', join('', @phrases), "|\n";
[download]

Output:

17:36 >perl 1667_SoPW.pl
|Web-developer, perl-program, element-explicit, function-call, 2-x spe
+ed.|

17:36 >
[download]

Hope that helps,