Congratulations! You win the prize for the Most-Confusing-Question Award:)

If I understand you correctly, this may get you started.

#! perl -slw
use strict;
use Data::Dumper;
sub rndStr{local $"=''; "@_[map{rand @_} 0 .. shift]"; } #!"

my @lines;
push @lines, "desc$_\tconid@{[int rand 10]}\t". rndStr(30, map chr, 32
+ ..255) for 1..100;

my (%lineChars, %totalChars, %conIDs);

for (@lines) {
    my ($descID, $conID, $string) = split /\t/;
    print "$descID, $conID, '$string'";
    push @{$conIDs{$conID}}, $descID;

    my %nonASCII;
    $nonASCII{$1}++ while $string =~ m[([^\x01-\x7f])]cog;
    $lineChars{$descID} = [ keys %nonASCII ];
    $totalChars{$_} += $nonASCII{$_} for keys %nonASCII;
}

print Dumper \%conIDs, \%totalChars, \%lineChars;
[download]

I've just generated some random data into an array to test it with.

The output of the for (which would probably be a while(<>) loop in your case), is three hashes.

• %conIDs which relates the each conID to an array of descID's it was associated with.
• %totalChars which gives the total counts for each non-ASCII char found in the strings
• %lineChars giving an array of the chars found in each string indexed by descID.

Of course it's quite probable that I've completely misread you and this is nothing like what your after, but it might suggest some ideas to you.