I was about to suggest decode_entities in HTML::Entities (which converts HTML/XHTML entities such as é to UNICODE characters), but while = looks like an HTML/XHTML entity, it isn't.

If you can't find a module that does what you want, you could model one based on HTML::Entities.

There are web sites that appear to offer listings of "standard" character entities, with the unicode code point values for each entity name -- I found one list of tables here, but you may need to look further to find others.

Given a reference like the one just cited, I might save each reference page to a file and use perl to convert it to a mapping table, like this:

my $entity;
while (<>) {
    if ( /<!ENTITY\s+(\S+)\s+"([^"]+)/ ) {
        my ( $name, $char ) = ( $1, $2 );
        $char =~ s/\&\#x([0-9a-f]{2,4});/chr(hex($1))/e;
        $char =~ s/\&\#([0-9]+);/chr($1)/e;
        $entity{$name} = $char;
    }
}

binmode STDOUT, ":utf8";
print "$_\t$entity{$_}\n" for ( sort {$entity{$a} cmp $entity{$b}} key
+s %entity );
[download]

Now I just need to save (redirect) the output of that process to a file, and use the file as a lookup table in any script that is going to convert character entity references to unicode characters. Just read that file into a hash (just like the %entity hash in the script above), and use the hash to filter data like this:

open( ENTS, "<:utf8", $entfile ) or die "$entfile: $!";
my %entity = map { split } <ENTS>;

my $enames = join "|", keys %entity;

while (<>) {
    s/\&($enames);/$entity{$1}/g;
}
[download]