diff --git a/tools/FILES b/tools/FILES
index 14cdeddd41..eabf9178dd 100644
--- a/tools/FILES
+++ b/tools/FILES
@@ -11,6 +11,7 @@ sample.emacs
 buildzip.pl
 romsizetest.pl
 wn2rdf.pl
+dict2rdf.pl
 make.inc
 makesrc.inc
 fwpatcher/*.[ch]
diff --git a/tools/dict2rdf.pl b/tools/dict2rdf.pl
new file mode 100755
index 0000000000..c9085d35ed
--- /dev/null
+++ b/tools/dict2rdf.pl
@@ -0,0 +1,83 @@
+#!/usr/bin/perl
+
+#             __________               __   ___.
+#   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
+#   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
+#   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
+#   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
+#                     \/            \/     \/    \/            \/
+# $Id$
+#
+# Copyright (C) 2005 Tony Motakis
+#
+# All files in this archive are subject to the GNU General Public License.
+# See the file COPYING in the source tree root for full license agreement.
+#
+# This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
+# KIND, either express or implied.
+
+# set the word size limit
+$word_limit = 32;
+
+use Compress::Zlib;
+
+# generate base 64 convertion hash
+@b64_values = (	'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L',
+		'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X',
+		'Y', 'Z', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j',
+		'k', 'l', 'm', 'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v',
+		'w', 'x', 'y', 'z', '0', 1, 2, 3, 4, 5, 6, 7, 8, 9, '+', '/' );
+
+foreach (0..63) {
+	$b64_get_value{$b64_values[$_]} = $_;
+}
+
+# base 64 convertion subroutine. note that if input is plain (base 64) 0, perl
+# doesn't like it, and the function misinterprents it as a (decimal) 0
+# while it actually is a (decimal) 52. Input has a tab in front anyway, so
+# this bug actually doesn't matter
+sub base64 {
+	my $i = 1, $num = 0, $left = $_[0];
+	while($left) {
+		$left =~ m{([^\s])$};		# use last char of string
+		chop $left;			# yes, chop, NOT chomp
+		$num += $i * $b64_get_value{$1};
+		$i *= 64;
+	}
+	$num;
+}
+
+# Open input files. <INDEX> is the database index, and $DICT is the actuall
+# dictionary file we want to access (note the use of zlib, hence the $DICT
+# variable instead of a <DICT> filehandle). <RDFOUT> is the output file, in
+# plain rockbox dictionary format
+open INDEX, $ARGV[0] or die "Could not open index: $!";
+$DICT = gzopen($ARGV[1], "rb") or die "Could not open definitions file: $!";
+open RDFOUT, ">$ARGV[2]" or die "Could not open output file: $!";
+
+# Read the index
+while(<INDEX>)
+{
+	next if /^00-?database/;
+
+	my @current = split /\t|\n/;			# split in pieces
+	$current[0] =~ s/^\s(.{1,$word_limit}).*$/\L\1/;	# lowercase
+	push @def_list, $current[0];
+	$def_begin{$current[0]} = base64($current[1]);
+	$def_length{$current[0]} = base64($current[2]);
+}
+
+# sort the definition list. input from the <INDEX> is usualy sorted, but this
+# is not mandatory in the dict file format, so we can't rely on this
+@def_list = sort @def_list;
+
+# read the whole DICT file into memory. overkill? propably. but the file is
+# compressed, and we need quick access to random parts of it
+$def_all .= $_ while($DICT->gzread($_));
+
+foreach (@def_list) {
+	$def = substr $def_all, $def_begin{$_}, $def_length{$_};
+	$def =~ s/\n\s*/ /g;	# remove newlines and whitespace after them
+	print RDFOUT $_ . "\t" . $def . "\n";
+}
+