koreader and stardic

[didina]

Dear all, I am re-posting here a question I have asked in a different section of the forum, but perhaps here it is more appropriate since koreader is much used for koboreader devices -

would you be interested in a stardic version of Latin-English inflected? I have been trying to convert it from a bgl file for several weeks with no success, it's because I have recently discovered the wonderful Koreader which does not flash on my e-ink device (icarus illumina) when looking for words and the Latin-English inflected would solve an enormous problem. So, I have tried to convert it in various ways, being no expert at all in such things, somehow I understood how to use the Python converter, I got the files, with a very big syn. file (around 21M) but Koreader does not recognize the dictionary, even if it does recognize all the other stardictionaries I could find around for other languages. I am posting here the bgl file and the stardic files I converted it into - they work nicely on Golden Dictionary and Color Dictionary, but not on Koreader. If anyone could help with this, I would be extremely grateful. It might be because I have not used Stardic Tools, they wouldn't install on my computer due to some dll missing; or perhaps there's some feature of Koreader that I don't know of. Or, the syn. file could somehow be integrated into the idx. file?

[Ken Maltby]

Markismus is our resident Koreader dictionary expert, he would be your best shot at a useful answer.

Luck;
Ken

[Markismus]

@didina
I've checked your dictionary with StarDict-Editor. It checks out o.k. When I de- and recompile it via xml (Textual StarDict dictionary) it gives only some small errors. I can decompile to a tabbed text-file, but it gives the error "Error, duplicate word a." when I try to recompile it.

Koreader uses sdcv, the commandline version of stardict. When I give the command

sdcv -l --data-dir .

it usually lists the name of the dictionary in the directory and the wordcount. For all dictionaries that I tried, it gives a nice reply. For your dictionary it remains empty.

So I googled "sdcv syn-files" and got the result that syn support for sdcv is still a feature request.

If that doesn't stop you, you could try to edit the tabbed text file generated by StarDict-Editor. It crashes on the recompilation of it, because of the multiple entries for a: interj./prep./prep.:

a <b>a</b><br/><font color="#457820">interj.</font> Ah!; (distress/regret/pity, appeal/entreaty, surprise/joy, objection/contempt);
a <b>a</b><br/><font color="#457820">prep.</font> ante, abb. a.; [in calendar expression a. d. = ante diem => before the day];
a <b>a</b><br/><font color="#457820">prep.</font> by (agent), from (departure, cause, remote origin/time); after (reference);
ab <b>ab</b><br/><font color="#457820">prep.</font> by (agent), from (departure, cause, remote origin/time); after (reference);
abactius <b>abactius, abactia, abactium</b><br/><font color="#457820">adj.</font> stolen/rustled (of cattle);
abactor <b>abactor, abactoris</b><br/><font color="#457820">n. m.</font> cattle thief, rustler; one who drives off;
abactus <b>abactus, abacta, abactum</b><br/><font color="#457820">adj.</font> driven away/off/back; forced to resign (office); restrained by; passed (night);
abactus <b>abactus, abactus</b><br/><font color="#457820">n. m.</font> cattle thieving, stealing of cattle, rustling;

Now what you read is: a | tab | whole lot of meaning | end-of-line.
If you remove the "end of line | a | tab" between the 1st and second line and the second and third line, you would end up with one entry for 'a'.

Now the new error would become "Error, duplicate word abactus."

So you could write a script to eliminate the multiple entries and recompile it with StarDict-Editor. However, you would still lose the function provide by the syn-files. Better look at Whitaker's Words here on mobileread before you decide that your dictionary is worth that extra effort.

[didina]

thank you very much. yes, I had kept googling and one of the things that crossed my mind was that the syn. file is the problem, but I wasn't sure since no one else seemed to have worried about it in connection with koreader. another member of the forum has kindly reconverted the bgl file with stardic tools and his version works on my device for the basic forms (just like Whitaker's Words, thanks for the suggestion). https://www.mobileread.com/forums/sho...88#post3121788
the inflected forms are not yet seen, I am still hoping it might be something with my icarus illumina, but the chances are low. I was wondering if it would not be possible to somehow modify the tabulated text version so as to make a separate entry for each inflected form. that would increase greatly the dimensions of the dictionary, but I think it might work. I have the stardic version of rosen's greek forms, which has around 900.000 entries, and that works without any problem. the inflected latin would have about 1.000.000 entries, by analogy it might work as well. Do you think I could do something to the tab txt file, or to the csv file, to make these separate entries? I am sorry, I am a complete ignorant in such stuff, this might a silly question, but I thought I should ask.

[Markismus]

@didina Just looking at these two entries:

abactus <b>abactus, abacta, abactum</b><br/><font color="#457820">adj.</font> driven away/off/back; forced to resign (office); restrained by; passed (night);
abactus <b>abactus, abactus</b><br/><font color="#457820">n. m.</font> cattle thieving, stealing of cattle, rustling;

First off, I know you can convert the dictionary so that you lose the formatting like "<font color=....>" and "</font>. I'll also add the <TAB> and <EOL> explicitly. The lines become simpler:

abactus<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night);<EOL>
abactus<TAB><b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>

Remember that sdcv only cares for the string in front of the <TAB> and the string between the <TAB> and <EOL>. All other code-like characters are hopefully not displayed in the output of sdcv.

Now two things need doing. First off: remove double entry:

abactus<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night); b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>

Then double the lines for the multiple forms. There are three in this example:

abactus<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night); b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>
abactum<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night); b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>
abacta<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night); b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>

There you go. Should be a nice scripting exercise to generate that!



I seem to remember something like:

abactus<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night); b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>
abactum<TAB><GOTO>abactus<EOL>
abacta<TAB><GOTO>abactus<EOL>

This would have a dramatic impact on the final size of the dictionary! However, I can't remember the code that I have symbolized as <GOTO>, nor do I know for sure that this isn't just the syn-file syntax all over again.

However, as koreader is ever developping, I do know that a feature to allow to search in the dictionary popup has been requested. Don't know whether it is implemented yet. That would get rid of the need of a coded <GOTO> and you could just code:

abactus<TAB><b>abactus, abacta, abactum</b><br/>adj. driven away/off/back; forced to resign (office); restrained by; passed (night); b>abactus, abactus</b><br/>n. m. cattle thieving, stealing of cattle, rustling; <EOL>
abactum<TAB>see: abactus<EOL>
abacta<TAB>see: abactus<EOL>

[Markismus]

For later reference the link to the stardict file-formats might be handy.

[didina]

thank you very much for your time and advice. I finally gave up and I bought a tablet, to use the golden dictionary with my texts. it does not feel alright, the e-ink was so much closer to real paper and less tiring. it would be great if koreader will at some point support syn files, I will continue to check the releases.

[filmo]

Some Onyx ereaders use Android with poor to good sucess. Read Onyx forum to stay current on their progress.

https://www.mobileread.com/forums/forumdisplay.php?f=220

[didina]

managed to convert the syn file using penelope. I've attached the result here. https://www.mobileread.com/forums/sho...36#post3128536

[Markismus]

So that new option in Penelope is useful! Very nice.

I've checked your dictionary with stardict editor and sdcv (v0.42).

$ sdcv -l --data-dir ./
Dictionary's name   Word count
Dictionary latininfl -> english    1223585

$ sdcv --data-dir ./ quaerit
Found 1 items, similar to quaerit.
-->Dictionary latininfl -> english
-->quaerit

<b>quaero, quaerere, quaesivi, quaesitus</b><br/><font color="#457820">v.</font> search for, seek, strive for; obtain; ask, inquire, demand;

Apparently the stardict editor is not so happy with your dictionary:

Verifing dictionary files...
Loading index file: /home/markismus/Downloads/stardict-dic/fulllatininflected[1]/latin infl+lewis.idx
Wrong order, first word = HS, second word = Habraea
Wrong order, first word = IIvirum, second word = Iacchus
Wrong order, first word = PRO, second word = Pactolus
Wrong order, first word = SS., second word = Sabaeus
Wrong order, first word = Zephyrus, second word = a
Incorrect number of words: in .ifo file, wordcount=1223585, while the real word count is 1427152
Failed!

A few tests show that the dictionary is not perfect, yet:

$ sdcv --data-dir ./ servus
Found 2 items, similar to servus.
0)Dictionary latininfl -> english-->cervus
1)Dictionary latininfl -> english-->nervus
Your choice[-1 to abort]: -1

sdcv --data-dir ./ canis
Found 1 items, similar to canis.
-->Dictionary latininfl -> english
-->canis

<b>canus, cana, canum</b><br/><font color="#457820">adj.</font> white, gray; aged, old, wise; hoary; foamy, white-capped; white w/snow/frost;

The dictionary can't find slave nor dog. If I query cane I get "canes, canis", so it is in the dictionary. It does find domus and frater.


So my conclusions are that you're onto something here, but it's not done yet. I would decompile your dictionary with stardict-editor to a tabbed text-file and compare the results from stardict-editor verify to it, run a lot of words through it to test it and see whether you can find a logic to its imperfect functioning.

If we know what's actually wrong, instead of just not-correct translation, we could create an issue on penelope's github page: https://github.com/pettarin/penelope/issues

[didina]

Apologies for replying late. Unfortunately I am very busy this period, I don't think I would be able to do this checking anyway. I noticed definitions go astray for longer forms, e.g. plus que parfait forms. I guess the problem is rather that one cannot move within the dictionary window from one inflected form (found in latin-english inflected or rosen's greek forms) to the basic dictionary entry in the larger dictionary (lewis for latin and liddle scott for greek). I recall that there was at some point the intention of enabling this feature within the dictionary window. But perhaps it would be too difficult to implement, again, in this technical matters I have very very little knowledge, I am anyway grateful to koreader developers for all they did.

[Markismus]

Koreader's chrox just implemented the feature hold to lookup word in dictionary window.

[didina]

that is great, I've checked it and it works, and it is really useful for latin. it does not work for ancient greek unfortunately, when tapping on a word inside the dictionary window it just selects a letter instead of the whole word.

[Ken Maltby]

Originally Posted by didina

that is great, I've checked it and it works, and it is really useful for latin. it does not work for ancient greek unfortunately, when tapping on a word inside the dictionary window it just selects a letter instead of the whole word.

This could be just a conflict of terms, but you select the word(s) by a long press/hold, not a tap as used most of the time. For multiple words it is a long press & slide across the words to be included.

Luck;
Ken

[didina]

thank you for your reply, I've checked and it still does select just a letter instead of the whole ancient greek word, even when holding on the word and not just tapping. the selection works when taking directly a word from the text, but not inside the dictionary window.