i want to distinguish between french word and english word, only do the decode, encode operation when it is french.

You want something like this, then:

s/%([0-9a-f]{2}/chr(hex($1))/egi;

if ( /[\x80-\xff]/ ) {
    push @new_words, encode( "iso-8859-1", decode_utf8( $_ ));
}
else {
    push @new_words, $_;
}
[download]

The point there is that you only need to do the encoding conversion if the string happens to contain any bytes with the 8th bit set (i.e. bytes in the numeric range 128-255).

Update: be aware that for this sort of approach, if the input data happen to contain any characters that are not in the iso-8859-1 table (e.g. certain "smart quote" characters, or Greek or Russian or ...), you'll get "?" instead of the intended characters as a result of the "encode(iso-8859-1)" call. That's just a limitation you have to live with if you have to stick with that old "legacy" iso-8859 encoding.