Use a constructed regular expression to perform the matching on the keys:

use strict;
use warnings;
use encoding 'UTF8';

my $aData = <<ADATA;
This is a boy.
&#12495;&#12531;&#12459;
She is a girl.
I am a man.
This apple is big.
That orange is small.
ADATA

my $bData = <<BDATA;
is a
&#12495;&#12531;&#12459;
small
BDATA

open my $inB, "<:utf8", \$bData;
my (@keys) = <$inB>;
close ($inB);
chomp @keys;

my $match = join '|', sort {length $b <=> length $a} @keys;
$match = qr{($match)};

open my $inA, "<:utf8", \$aData;
while (defined (my $line = <$inA>)) {
    print "Found '$1' in line $.:\n   $line" if $line =~ $match;
}
[download]

Prints:

Found 'is a' in line 1:
   This is a boy.
Found 'ハンカ' in line 2:
   ハンカ
Found 'is a' in line 3:
   She is a girl.
Found 'is a' in line 5:
   This apple is big.
Found 'small' in line 6:
   That orange is small.
[download]

Note: I don't seem to be able to paste ハンカ into the code blocks. Replace the entities with the actual string for correct behaviour.