Pocketbook dictionary

[tropoy]

Originally Posted by EastEriq

https://www.mobileread.com/forums/sh...17&postcount=4

Thanks a lot

But I spoke about "Le Nouveau Littré". It means the new "Littré". The "Littré" is free but "Le Nouveau Littré" is not.
So, I wonder if it's possible to recover the one that is inside my Vivlio.

Thanks

[Markismus]

Originally Posted by tropoy

Hi,
Is anybody has recover the "Nouveau Littré" Dictionnary from the Inkpad 3 (Vivlio) ?
I bought one Nouveau Littré few years ago, but I have two files : A ".dict" and a ".dict.idx". But it's not compatible with Inkpad 3. :/
Thanks

If it is Stardict format, I would also expect an “.ifo” file. Why don’t you upload the files to a file server and post a link? Or install a dictionary converter like Linguae or Penelope and try to import and convert it to Stardict or dxdf-format. From either you can convert to dic-format.

[nhedgehog]

Originally Posted by tropoy

Thanks a lot

But I spoke about "Le Nouveau Littré". It means the new "Littré". The "Littré" is free but "Le Nouveau Littré" is not.
So, I wonder if it's possible to recover the one that is inside my Vivlio.

Thanks

If the dictionary is still on your Vivlio, you can recover it via bash script or PBTerm (though it may be encrypted, some dics are and some not).

[tropoy]

Originally Posted by Markismus

If it is Stardict format, I would also expect an “.ifo” file. Why don’t you upload the files to a file server and post a link? Or install a dictionary converter like Linguae or Penelope and try to import and convert it to Stardict or dxdf-format. From either you can convert to dic-format.

That's what I understood about the Stardic, I miss this ".ifo" File.

I tryed to use Penelope without success (maybe I didn't installed Python correctly).

I putted my files here :
https://1fichier.com/?m2moq80o797u2wgvytws

Someone could convert them into a Pocketbook compatible format ?

Thanks in advance

[Markismus]

TLDR: The dictionary is not encrypted nor binary. Conversion will probably take a little programming, but not much. Maybe someone recognizes which dictionary format uses the described structure (sqlite3 database for index and zipped textfiles for meaning) and you can use that knowledge to use a converter.

I downloaded them and it appears that the dict-file is actually an archive file containing markup text in 120 files numbered c_1 to c_120. The idx-file is actually a sqlite3-database that can be browsed with any sqlite browser. (I used sqlitebrowser.)

In the database there are 4 tables. One table is called T_DictIndex and holds per row a Word, an Offset, a Size and the Chunk number:

F_Word	F_Offset	F_Size	F_ChunckNum
a	0		574	1

In chunk c_1, the first 564 characters are:

<b>A,</b> N.&nbsp;m. [<f>&a;</f>] ou [<f>&â;</f>] Voyelle et première lettre de l'alphabet. <i>Une panse d' <i>a </i>, </i>la première partie d'un petit <i>a </i> dans l'écriture. <f>&os;</f>&nbsp;<i>N'avoir pas fait une panse d' <i>a </i>, </i>c'est-à-dire n'avoir rien écrit. <f>&ns;</f>&nbsp;<i>Prouver par A + B, </i>avec précision et rigueur. <f>&ns;</f>&nbsp;<i>De A à Z, </i>du début à la fin. <f>&ns;</f>&nbsp;<i>A4, </i>format d'une feuille de papier de 21&nbsp; X &nbsp;29,7&nbsp;cm. <i>A3, </i>format 29,7&nbsp; X &nbsp;42&nbsp;cm. <f>&ns;</f>&nbsp;La.

F_Word	F_Offset	F_Size	F_ChunckNum
à	574		2523	1

In chunk c_1, the next 2444 chars are:

<b>À,</b> Prép. [<f>&a;</f>] (lat. <i>ad</i>, mouvement, direction, proximité, lat. <i>ab</i>, séparation, origine, et lat. <i>apud</i>, relation, accompagnement) <f>&os;</f>&nbsp;Marque direction, tendance&nbsp;: <i>aller à Rome &nbsp;; aimer à lire. </i>&nbsp;<f>&os;</f>&nbsp;S'emploie devant le régime indirect des verbes actifs&nbsp;: <i>donner de l'argent à un pauvre. </i>&nbsp;<f>&os;</f>&nbsp;Sert à déterminer le lieu où est quelque chose, où s'exécute une action&nbsp;: <i>résider à Paris &nbsp;; être à sa place. </i>&nbsp;<f>&os;</f>&nbsp;Sert à indiquer le temps, le moment, etc.&nbsp;: <i>à la fin du mois. </i>&nbsp;<f>&os;</f>&nbsp;Marque appartenance, possession&nbsp;: <i>rendez à César ce qui est à César &nbsp;; il a un style à lui. </i>&nbsp;<f>&os;</f>&nbsp;Avec un complément indique l'espèce&nbsp;: <i>vache à lait &nbsp;; </i>la qualité&nbsp;: <i>or à vingt-deux carats &nbsp;; </i>la forme ou la structure&nbsp;: <i>clou à crochet, table à tiroir &nbsp;; </i>la destination&nbsp;: <i>marché à la volaille &nbsp;; </i>la conformité, la convenance&nbsp;: <i>à mon avis &nbsp;; </i>l'instrument&nbsp;: <i>pêcher à la ligne &nbsp;; </i>la mesure, le poids, la quantité&nbsp;: <i>vendre à la livre, à la douzaine &nbsp;; </i>le prix, la valeur&nbsp;: <i>pain à vingt centimes la livre, dîner à trois francs &nbsp;; </i>l'intention&nbsp;: <i>à regret &nbsp;; </i>la cause&nbsp;: <i>se ruiner à jouer &nbsp;; </i>l'effet, le résultat&nbsp;: <i>blesser à mort. </i>&nbsp;<f>&os;</f>&nbsp;<i>À </i> précédé et suivi du même mot marque succession, gradation, ordre&nbsp;: <i>deux à deux &nbsp;; </i>jonction&nbsp;: <i>bout à bout &nbsp;; </i>opposition&nbsp;: <i>face à face. </i>&nbsp;<f>&os;</f>&nbsp;<i>À </i> se place après certains adjectifs pour en déterminer le sens&nbsp;: <i>fa cile à dire &nbsp;; prêt à combattre. </i>&nbsp;<f>&os;</f>&nbsp;<i>À </i> suivi d'un infinitif équivaut souvent au participe précédé de <i>en </i>&nbsp;: <i>à vrai dire. </i>&nbsp;<f>&os;</f>&nbsp;<i>À </i> devant un infinitif peut quelquefois s'expliquer par<i>de quoi </i>&nbsp;: <i>verser à boire. </i>&nbsp;<f>&os;</f>&nbsp;<i>À </i> indique ce qu'on doit faire&nbsp;: <i>c'est un avis à suivre &nbsp;; </i>ce qui doit être la suite d'un événement&nbsp;: <i>c'est une affaire à vous perdre. </i>&nbsp;<f>&os;</f>&nbsp;<i>À </i> s'emploie dans certaines phrases elliptiques&nbsp;: <i>à moi &nbsp;! au feu &nbsp;! à ta santé &nbsp;! </i>

So the way the index count is not characters, but probably bytes. There are a lot of   and &os; that only code for one symbol and there are a lot of accented characters that have higher unicodes, so are coded by more than 1 byte.

All in all, it's close enough to see the pattern: The index words are surrounded by <b> bold start and </b> stop tags. However, the description also holds start and stop tags.

So if you
(1) take the index words from the table and
(2) search for the text between two following index words and
(3)put the first index word and found text separated by a delimiter such as a comma on a line and
(4) repeat that for all 69389 entries,
than you have reconstructed your dictionary in Stardict csv-format, which can be converted to pocketbook format.

You could also use the found terms and place them directly in a xdxf-format. From xdxf-format you can use pocketbooks converter.exe to convert it to pocketbook's dic-format.

[Markismus]

From Penelope it seems to be a Bookeen dictionairy.

I used penelope to convert Bookeen to Stardict:

 python -m penelope -i NouveauLittre.dict,NouveauLittre.dict.idx -j bookeen -f fr -t fr -p stardict -o output

Then I used my script from Pocketbook dictionary format revisited to convert it to pocketbook dictionary format that is at least compatible with PB740-2. (Inkpad 3 Pro).

I've uploaded the results* to a fileserver and attached them.

_____________________
*These are actually the tested and corrected results from a few posts down.

[tropoy]

Originally Posted by Markismus

From Penelope it seems to be a Bookeen dictionairy.

I used penelope to convert Bookeen to Stardict:

 python -m penelope -i NouveauLittre.dict,NouveauLittre.dict.idx -j bookeen -f fr -t fr -p stardict -o output

Then I used my script from Pocketbook dictionary format revisited to convert it to pocketbook dictionary format that is at least compatible with PB740-2. (Inkpad 3 Pro).

I've uploaded the results to (...)

Wow, I admit that I did not understood much of your #275 post. For me it seem to be a hard work.
But I thank you a lot for having taking the time to convert my dictionnary.
I'll try it ASAP in my Bookeen Inkpad 3.
Thanks again.

[Markismus]

Originally Posted by EastEriq

https://www.mobileread.com/forums/sh...17&postcount=4

Originally Posted by tropoy

Wow, I admit that I did not understood much of your #275 post. For me it seem to be a hard work.
But I thank you a lot for having taking the time to convert my dictionnary.
I'll try it ASAP in my Bookeen Inkpad 3.
Thanks again.

My file browser (Thunar) identified the file types after I downloaded them. The hard work was all in typing the post and creating the script (, which I already wrote last month.)

I did forget to test the dictionary, because the generated Stardict info-file looked fine. If something is off about the dictionary, would you please let me know?

If others would like their dictionaries converted, too: Post in the thread Pocketbook dictionary format revisited

[EastEriq]

Originally Posted by Markismus

If something is off about the dictionary, would you please let me know?

Only since you're saying. I've looked at it, and there are still so many tags and nonstandard entities which make it poorly readable. They should be all workable out with some scripting. I started writing some sed substitutions for the xdxf, but at some point I gave up. Among them:

• html entities like &nbsp; &apos; &# 171; &# 187; to be replaced with their unicode equivalent;
• coded characters. The ones I identified were greek letters, e.g. &#x03B5; for ε. To be replaced by the unicode.
• phonetic/splitting patterns, like [<f>&a;&b;&è;&s;(&e;)&m;&an;</f>] to be replaced with something proper e.g. [a-b-è-s-(e)-m-an] (be careful because there are tricky cases w.r.o doubles, silent vowels, pronunciation equivalents, etc.)
• non entity splitters, like <f>&os;</f>, <f>&ns;</f>, <f>&oo;</f>. I think they separate different meanings of a lemma or explanation from example. To be replaced by a proper graphism.

[Markismus]

Yes, I see it. Garbage in = garbage out, I am afraid.

This is the entry in the chunk c_1 in the archive dict-file:

<b>A,</b> N.&nbsp;m. [<f>&a;</f>] ou [<f>&â;</f>] Voyelle et première lettre de l'alphabet. <i>Une panse d' <i>a </i>, </i>la première partie d'un petit <i>a </i> dans l'écriture. <f>&os;</f>&nbsp;<i>N'avoir pas fait une panse d' <i>a </i>, </i>c'est-à-dire n'avoir rien écrit. <f>&ns;</f>&nbsp;<i>Prouver par A + B, </i>avec précision et rigueur. <f>&ns;</f>&nbsp;<i>De A à Z, </i>du début à la fin. <f>&ns;</f>&nbsp;<i>A4, </i>format d'une feuille de papier de 21&nbsp; X &nbsp;29,7&nbsp;cm. <i>A3, </i>format 29,7&nbsp; X &nbsp;42&nbsp;cm. <f>&ns;</f>&nbsp;La.

And this is the entry in the Stardict file generated by Penelope:


As you can see, it is not the accented characters that are the problem, rather in the idx-file there is a doctype definition given, which defines all those entities:

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"  "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"[
<!ENTITY ns "♦">
<!ENTITY os "•">
<!ENTITY oo "›">
<!ENTITY co "‹">
<!ENTITY a  "a">
<!ENTITY â  "&#x0251;">
<!ENTITY an "&#x0251;&#x303;">
<!ENTITY b  "b">
<!ENTITY d  "&#x0257;">
<!ENTITY e  "&#x0259;">
<!ENTITY é  "e">
<!ENTITY è  "&#x025B;">
<!ENTITY in "&#x025B;&#x303;">
<!ENTITY f  "f">
<!ENTITY g  "&#x0261;">
<!ENTITY h  "h">
<!ENTITY h2 "&#x0027;">
<!ENTITY i  "i">
<!ENTITY j  "J">
<!ENTITY k  "k">
<!ENTITY l  "l">
<!ENTITY m  "m">
<!ENTITY n  "n">
<!ENTITY gn "&#x0272;">
<!ENTITY ing "&#x0273;">
<!ENTITY o  "o">
<!ENTITY o2 "&#x0254;">
<!ENTITY oe "&#x0276;">
<!ENTITY on "&#x0254;&#x303;">
<!ENTITY eu "&#x0278;">
<!ENTITY un "&#x0276;&#x303;">
<!ENTITY p  "p">
<!ENTITY r  "&#x0280;">
<!ENTITY s  "s">
<!ENTITY ch "&#x0283;">
<!ENTITY t  "t">
<!ENTITY u  "&#x0265;">
<!ENTITY ou "u">
<!ENTITY v  "v">
<!ENTITY w  "w">
<!ENTITY x  "x">
<!ENTITY y  "y">
<!ENTITY z  "z">
<!ENTITY Z  "&#x0292;">]>
<html xml:lang="fr"
	xmlns="http://www.w3.org/1999/xhtml">
	<head>
		<title></title>
	</head>
	<body>

However, the &nbsp sequence is normal html for non-breakable-space. If you change the sametypesequence in the ifo file from m to h they don't disappear from linguae, but Goldendict does switch (as does Koreader):

sametypesequence=m



sametypesequence=h

[tropoy]

Originally Posted by Markismus

(...)
I did forget to test the dictionary, because the generated Stardict info-file looked fine. If something is off about the dictionary, would you please let me know?

If others would like their dictionaries converted, too: Post in the thread Pocketbook dictionary format revisited

Hi,

I've tested the dic file in my inkpad 3.

It work fine if I use it as I am reading a book and I press on a word. I have the definition and the layout of the page is good.

But, if I use the dictionary directly in the dictionary manager, I lose the layout and some strange characters are inserted in the text, as "&nbsp;<f>&os;</f>&nbsp" by example. That's what mentioned EastEriq.
And as said EastEriq, the phonetic paterns are not present when reading a book and separated by "&" in the dictionary manager.

I think it's strange that it works perfectly while I read a book and not in the dictionary manager.
But I rarely use the dictionary manager...

[Markismus]

@tropoy, it seems that your dictionary program used whilst reading a book properly handles html automatically and knows the apparently typical symbol references defined in the doctype. This is somewhat expected, because converter.exe seems to strip out all html-tags that are more complex than bold and italics: All the interpretation needs to be on the interpreter side.

However, your dictionary manager doesn't know html at all, it seems. What dictionary manager are you referring to?

@EastEriq The tags <f> and </f> are new and were demolished by my script with the conversion of ">" to "&lt;". This is corrected now in the script. Also now all the ampersands are left alone if they are closely followed by a ";".

[Markismus]

Would you test these, please?

I've created a filter to replace all instances of a named entity with its numerical counterpart based on a given DOCTYPE.

The current information flow is:
Bookeen dict -->Using Penelope--> Stardict-->using script --> xdxf
xdxf -->using script-->Pocketbook dic
xdxf -->using Linguae with the settings unsorted and sametypesetting h -->Pocketbook dic
(Linguae makes a mess of this dictionary if it tries to ASCII sort and searching for 'a' would then return 'ABN'.)

[tropoy]

Originally Posted by Markismus

However, your dictionary manager doesn't know html at all, it seems. What dictionary manager are you referring to?
.

I refered to the dictionary manager of my Inkpad 3. When you goes in the parameter tab and choose "dictionary".

I 've tryed your new files...

I't seems to make exactly the same as before. Always "<f>&os;</f>&nbsp" in the internal dictionary manager of the Inkpad3. And a good layout but no phonetic when reading a book.

Thanks

[EastEriq]

Ah, that there is an entity translation table somewhere makes sense. However, no joy with your latest .dic for me. Three different applications, pbreader, stock PB dictionary app and CR3 differ in whether they handle standard html entities like &nbsp; or not, and whether they literally render the custom ones or ignore them, but it seems they are all still there. Are you sure you uploaded your last conversion?