I think I wouuld reverse the order of the indexes, $lang first, so you can take a reference to the whole hash of all translations for one language. I do assume that you actually want to use just one language for the output, and this way, you can drop the reference to the language itself throughout the code. Yes, I would use hashes, and string indexes, in order to get something you can read. A bit like:

$translat{English}{red} = 'red';
$translat{English}{blue} = 'blue';
$translat{English}{yellow} = 'yellow';
$translat{Spanish}{red} = 'rojo';
$translat{Spanish}{blue} = 'azul';
$translat{Spanish}{yellow} = 'amarillo';
[download]

$t = $translat{Spanish};
print "The proper word for red is '$t->{red}'\n";
[download]

The proper word for red is 'rojo'
[download]

	English	Spanish	French
red	red	rojo	rouge
blue	blue	azul	bleu
yellow	yellow	amarillo	jaune

The first row is the languages, the first column the hash keys. You might think that it's redundant with the English word, but this is a way to distinguish between synonyms, the hash key doesn't have to be the same as the English word, for example with a word "blue" having more than one meaning, you can use "colour_blue" as the hash key for the colour.

Back to your question... You can combine strings into one hash key, for example "colour_0" for red, "colour_1" for blue, and "colour_2" for yellow, thus you can do your lookup using $t->{"colour_$colour"}. Maintainable and flexible. I like that.

Addendum: Some Perl code to read such a tab delimited text file into a HoH as described:

$_ = <TABLE>;
chomp;
(undef, my @lang) = split /\t/;
while(<TABLE>) {
    chomp;
    my($key, @text) = split /\t/;
    for my $i (0 .. $#lang) {
        $translat{$lang[$i]}{$key} = $text[$i];
    }
}
[download]