Using a hash will speed things up, but unless you are using some persistant environment (mod_perl etc), then reloading the hash everytime will still be quite slow.

If moving your dictionary into a database is not immediatly possible for any reason, then you can speed things up a lot by simply splitting it into several files based on the first letter, so that you have 26 files named dict_A.txt, dict_B.txt, etc. That would (on average) make your program take less than 5% of the time currently takes:

if (length($input)>0){
    print "<p><b>".$input."</b></p>";
    
    my $firstLetter = substr( $input, 0, 1 );

    open DICTE, "dict" . $firstLetter . ".txt" or die $!;

    while (<DICTE>){
        chomp;
        ($english, $lang, $irreg, $clss) = split(/;/,$_); 

        if ($input eq $lang){ 
            print "<p>$english - $lang, $clss</p>";
        } 
        if (conj("$lang;present;1;singular") eq $lang){ 
            print "<p>$english - $lang, $clss</p>";
        } 
        if (conj("$lang;present;2;singular") eq $lang){ 
            print "<p>$english - $lang, $clss</p>";
        } 

    }
}
[download]

And if that is still too slow, then you can divide the dictionary into smaller pieces by using the first two letters. dict_AA.txt, dict_AB.txt etc. which (on average) should reduce the search time to ~ 0.2% of the current.

Though there are probably few words begining with ZZ... or JJ... etc, so the speed up won't be exactly evenly distributed.

Moving your data into a simple tied DB like BerkeleyDB or similar, is a better long term solution, but this might help you out while you get to grips with Perl, hashes, modules and the rest.