Hi Ramon, I do not have source files for dawg dictionaries and I am not able to "decompile" them. Anyway I think to create dictionaries is the easiest part of tesseract training: based on wiki[1] input is simple utf-8 file with one word per line. This file is split to several files:
* lang.punc -> words with punctuation patterns
* lang.number -> words with number patterns
* lang.freq -> frequent words
* lang.word -> rest of the words
I believe you can get list of words from other opensource projects (e.g.
spellchecker, dictionary projects as apertium.org, or search for free
Catalan Corpus - do not forget to clear license of data first!) or you
can create it from wikipedia[2].
dawg files are easy to create (big input file can cause a long run this
command!):
$ wordlist2dawg [-t] word_list_file dawg_file unicharset_file
e.g. wordlist2dawg lang.punc lang.punc-dawg lang.unicharset
This command is valid for tesseract 3.00. wordlist2dawg in tesseract
2.04 do not use unicharset_file as input.
I hope there will be more details soon on
http://www.sk-spell.sk.cx/tesseract-ocr-en.
[1] http://code.google.com/p/tesseract-ocr/wiki/TrainingTesseract
[2] http://wiki.apertium.org/wiki/Building_dictionaries
Zdenko
Dn(a 29.04.2010 09:30, Ramon wrote / napísal(a):
> Hi for you quick answer Zdenko.
>
> As you pointed out, I'm already using tif / box pair from spanish
> language to train my catalan .traineddata language. (As spanish
> characters suits catalan characters too).
>
> But doing just this (with no words in dictionary files) the dictionary
> is not quite good. I think the difference is from the words used in
> those dictionaries. So I'm asking for that utf8 files...
>
> Don't know if you (or a developer) can provide them.
>
> Thanks.
>
> Ramon.
>
>
>
>
> On 28 Abr, 15:55, zdenko podobny <[email protected]> wrote:
>
>> Hello Ramon,
>>
>> for extending existing language you need "Tif/Box pairs"
>> seehttp://code.google.com/p/tesseract-ocr/wiki/FAQand there "How do I add
>> just
>> one character or one font to my favourite language, without having to
>> retrain from scratch?"
>>
>> Unfortunately tif/box pairs are provided only for eng, deu, fra, ita, nld
>> and spa languages... So you can wait that somebody will someday release
>> tif/box pairs for your language or you will start training from scratch. I
>> choose second option and this is reason why I started with testing of
>> training process for tesseract 3.00.
>>
>> BR,
>>
>> Zdenko
>>
>>
>>
>>
>>
>> On Mon, Apr 26, 2010 at 11:06 AM, Ramon <[email protected]> wrote:
>>
>>> Hi,
>>> After some tests I realized the best for me is to put effort to extend
>>> the Catalan Diccionari which is in svn repository (v3).
>>> It will be so useful if you can do one of these:
>>>
>>
>>> -> deliver the different files combined to create the cat.traineddata
>>> unified file. (the utf8 files used to generate the dawg would be also
>>> amazing!).
>>> -> show how to extract these files from the cat.traineddata and how to
>>> dawg2utf8 (if it is possible).
>>>
>>
>>> THANKS!
>>>
>>
>>> --
>>> You received this message because you are subscribed to the Google Groups
>>> "tesseract-ocr" group.
>>> To post to this group, send email to [email protected].
>>> To unsubscribe from this group, send email to
>>> [email protected]<tesseract-ocr%2bunsubscr...@goog
>>> legroups.com>
>>> .
>>> For more options, visit this group at
>>> http://groups.google.com/group/tesseract-ocr?hl=en.
>>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "tesseract-ocr" group.
>> To post to this group, send email to [email protected].
>> To unsubscribe from this group, send email to
>> [email protected].
>> For more options, visit this group
>> athttp://groups.google.com/group/tesseract-ocr?hl=en.
>>
>
smime.p7s
Description: S/MIME Cryptographic Signature
