Some modules on CPAN that you may wish to look into include String::Approx, Text::Levenshtein, Text::WagnerFischer, and Text::PhraseDistance. Good luck!

#!/usr/bin/perl

use strict;
use Data::Dumper;

my (@words, $substr, %hash);

@words = qw/sem sema cat semok hgt cato seme cate/;

foreach (@words) {
    $substr = substr( $_, 0, 3 );
    push( @{ $hash{$substr} }, $_ );
}

print Dumper(\%hash);

exit;


$VAR1 = {
          'cat' => [
                     'cat',
                     'cato',
                     'cate'
                   ],
          'sem' => [
                     'sem',
                     'sema',
                     'semok',
                     'seme'
                   ],
          'hgt' => [
                     'hgt'
                   ]
        };
[download]

But if your query is as simple as only sorting to first 3 letters, then the following may do the trick:

  my @words = qw(sem sema cat semok hgt cato
                 seme cate);
  my %similar = ();
  foreach my $word ( @words )
  {
      push( @{$similar{substr($word,0,3)}}, $word );
  }

  foreach my $prefix ( sort keys %similar )
  {
      print( join( ', ', @{$similar{$prefix}} ) );
      print( "\n" );
  }
[download]

cat, cato, cate
hgt
sem, sema, semok, seme
[download]

grinder's most excellent Regexp::Assemble can certainly help you with part of your problem, identifying bits common to multiple words. Take a look at Why machine-generated solutions will never cease to amaze me for a sample of what this module can do; you'll be impressed.

Oh, and grinder's scratchpad too...

my $match = sub { return substr($_[0], 0, 3) eq substr($_[1], 0, 3); }
+;
print $match->('aaaa','aaab');
[download]

myset($match, @words);
sub myset {
    my $match = shift;
    # Do whatever set algorithm here,
    # using $match to compare words
}
[download]