release
dev builds
extras
themes manual
wiki
device status forums
mailing lists
IRC bugs
patches
dev guide



Search | Go
Wiki > Main > RockboxDictionary (r21)

Dictionary

Note: this is still a work in progress, developers only.
Note2: If you can write a xxx2rdf script in perl, please do so. The more sources the better.

Todo

Name Progress Description
searching done Binary search in the dict plugin
rdf2binary done Conversion from the rockbox dictionary format to the internal binary format
wordnet2rdf done Conversion from the WordNet format to rockbox dictionary format
displaying 50% Wrapping lines is done, but we still need scrolling for long descriptions
interface 10% A better interface for the plugin, like a button for "new search" and things like that. (We now just have an exit button.)
xxx2rdf 0% Conversion from other formats to the rockbox dictionary format
fileformat 0% Improving the binary file format, more info below

The most interesting xxx2rdf would be the Dict format, as there are a lot of free dictionarys availible in that format.

Creating a dictionary file

1. Download the prolog version of the WordNet dictionary here: WNprolog-2.0.tar.gz

2. Extract wn_g.pl and wn_s.pl from it.

3. Put wn2rdf.pl, wn_g.pl, wn_s.pl and the rdf2binary tool in 1 directory, and execute wn2rdf.pl

4. Execute the rdf2binary tool, it will output dict.desc and dict.index

5. Copy dict.desc and dict.index to ./rockbox/rocks/apps on the player.

The rockbox dictionary format

The input format for rdf2binary is very simple at this moment. It's one line per word, starting with the word, then a tab and then the description. The only thing you should be aware of when creating this files is that they must be in alphabetical order, and all words should be in lowercase.

The binary format

The binary format used for the index is pretty simple, the struct is like this one:

struct {
    char word[WORDLEN];
    long offset;
};

WORDLEN is a define in the rdf2binary tool, and the plugin. And the offset is an offset in dict.desc where the description is stored.

The improved binary format

This is still an idea under construction, but the new format would be just 1 file containing:

Field Size Description
magic number 4 bytes A simple identifier to know if this is a valid file.
version 4 bytes A format version number, this way we can detect old files and return an error.
max_wordlen 4 bytes The maximum word lenght used in this file.
wordcount 4 bytes The word count.

After that there should be the index data:
Field Size Description
offset 4 bytes The offset from the beginning of the file to the description
word max_wordlen The word, lenght from the header.

And then just plain text description data, one description per line.

The hash binary format

Header:
Field Size Description
magic number 4 bytes A simple identifier to know if this is a valid file.
version 4 bytes A format version number, this way we can detect old files and return an error.
wordcount 4 bytes The word count.

Offset table:
Field Size Description
offset 4 bytes The offset from the beginning of the file to a value in the hash table

Hash table:
Field Size Description
name variable The word
offset 4 bytes The offset from the beginning of the file to the description

When searching for a word with hash X, the plugin looks up the offset for X and X+1 in the offset table. It reads the data between those offsets on looks for the word, we were searching for. It's just a hash table with chaining.

Sources for dictionary files

There is nearly everything needed for a german<->english single word translator on http://dict.tu-chemnitz.de
They are providing the wordlist under the GNU GPL Version 2 (whatever it means to have versions for licenses. Thats totally new to me). See http://ftp.tu-chemnitz.de/pub/Local/urz/ding/de-en/ to download a 6 MB text file of words and words and words smile

For those who don't have to tools for comiling their own Dictionary files, you can download them from http://www.rockbox.dreamhosters.com/dict.zip (6.6MB) . If you want to download the two parts separately just get, http://www.rockbox.dreamhosters.com/dict.desc (17MB) and http://www.rockbox.dreamhosters.com/dict.index (5.0MB).

-- PeterOlson - 12 Nov 2006
Edit | Attach | Print version | History: r24 | r22 < r21 < r20 < r19 | Backlinks | View wiki text | More topic actions...
r21 - 17 Aug 2007 - 16:59:45 - MichaelDiFebbo
Copyright by the contributing authors.