New feature: Tag translation

Blog post here: http://www.librarything.com/blogs/librarything/2012/06/new-feature-tag-translati...

Can you turn it off? Yes, of course. Every tag cloud has a link to turn it off, on or on with coloring.

Bad translations: There are definitely bad translations out there, coming from Wikipedia. Examples.

* Tag "classics" ( http://www.librarything.de/tag/classics ) is translated like the field of study (eg., Klassische Altertumswissenschaft http://www.librarything.de/tag/classics in German ). This is sometimes the meaning in English too, but not usually.
* Tag "twilight" ( http://www.librarything.de/tag/Twilight ) is translated literally, which isn't used in some languages (eg., Crépuscule http://www.librarything.fr/tag/Twilight in French )
* Tag "pet" ( http://www.librarything.de/tag/pet ) got coopted by the Wikpedia redirect for "PET" ("Positron emission tomography"), making most languages translate it as PET.
* Pages that linked to disambiguation pages. I'm currently getting rid of these.

Voting: Voting, both on tag pages and on the helpers page ( http://www.librarything.com/helpers_tagtranslations.php ) is a bit odd. Members are encouraged to vote good translations up and bad translations down. But votes have been pre-populated too, to give Wikipedia tags some votes. Also, all tags get extra votes for meeting various criteria. Here are the rules currently.

+1 if source is Wikipedia
+1 if translation was already within the "combined tags" for the tag.
+1 if the Wikipedia page's English title was the same as the winning English tag (ie., not the same as a tag alias)
-1 if tag contains parentheses (designed to hurt "disambiguation" pages, may go away if we can kill those)

How does this affect tag combination? It doesn't. Tags should still be combined. This does away with one possible argument against tag combination, namely that tag-combination subordinates non-English tags to English-tags. Now it doesn't need to.

Because we've now got so much tag-translation data from Wikipedia, I will be making some sort of tool to find tag-translations that suggest combinations. (If you get that, you get it. If you don't, it's hard to explain.)

Why: We've had plans to do tag translation, and the user interface for it, kicking around for a year or more. We've discussed it before as a future thing. I hope it makes LibraryThing feel more "natural" to new visitors, since it was perhaps the most prominent site element that was totally untranslated.

To be perfectly frank, however, the impetus to push this now, and spend a few days finishing it, is that we have some major German customers for LibraryThing for Libraries and they're getting blow-back about English tags. LibraryThing for Libraries is doing surprisingly well overseas, but this is definitely a stumbling block for it.

ENGLISH translations. On another thread (http://www.librarything.com/topic/138298) a member noted tht LibraryThing has a number of mutli-language tags where English isn't "winning." Shouldn't LibraryThing members be able to translate those tags into English.

The short answer is: Yes. But it may not be done through tag translation, but rather through the canonization process used in CK. I haven't yet decided, but the goal is a good, certainly.

Will we be able to translate tags into English where the default -- even on the English-language .com site -- is in a foreign language? If not, is there some other action we can take?

I ask because the way tag translation works at the moment we're translating tags from English to other languages, and that assumes that tags are in English to begin with. Not all of them do, even when combined with the English tags. I've pointed out that "in French" points to "en français. BarkingMatt wondered about "Arquitetura Moderna" which includes the English "modern architecture." Unless these tags and others like them -- those in which the not-in-English version 'wins out' in combining -- this translation thing will be, at best, of variable usefulness.

Sorry, I added a brief reply at the end of my first message. Short answer: Yes. Let's think about it a bit, though.

I think one ground rule would have to be that the tag's getting its non-English word from non-English speakers. That is, I don't think bildungsroman should be translated, because it's a technical term in English.

I'm also not convinced that I can't solve the problem technically. "Modern architecture" is going into Spanish because of a single user. Probably when deciding the final form of a combined tag, I should balance the total number of tags applied and the total number of users using them. That is, although "arquitectura modern" has the highest total number, it's coming from only one user.

Why won't that helpers list change language? I can only vote on the proposals for German.

Sorry. Changing now.

Why are the links in the original post to .de?

Could the default language to translate into be set to the one of the site you're on? I'm on se.librarything.com, so it's plausible that I'd enter more translations into Swedish than French.

Also, I tried to enter a translation, but nothing happened.

Could the default language to translate into be set to the one of the site you're on? I'm on se.librarything.com, so it's plausible that I'd enter more translations into Swedish than French.

Fixed. It uses your language now.

Why are the links in the original post to .de?

I was illustrating specific things. What I was saying might not apply to other languages.

Also, I tried to enter a translation, but nothing happened.

Hmm. Anyone else. I just succeeded in adding one on the Turkish site.

http://tr.librarything.com/tag/bunnies

I've been successfully adding translations on the French site.

Added a section for "Top helpers" on right of http://www.librarything.com/helpers_tagtranslations.php

Hmm... I think we might need threads to request help in voting down bad translations -- automatic or otherwise. It could include the "please vote against my bone-headed suggestion" notion seen in the 'Combiners!' group but also ask for help voting down bad auto-translations*. maybe there could be an additional thread asking for a "stupidity check" on translations, since I suspect the number of people with native fluency in multiple languages is fairly small -- even on a site as ostensibly literate as this one.

Only question is, what group? Or should we start one?

----------
* The automatic French translation for "New York" is "État de New York" -- a translation that really belongs over here. Looks like a few more translations need to be "moved" as well.

> 10

I think we might need threads to request help in voting down bad translations -- automatic or otherwise.

Agreed.

I'd think either the Taggers group or the Combiners! group. I lean towards the latter, as the threads for "vote yes"/"vote no" on tag combination are already there, and I think that group is more inclined to engage in these sorts of projects.

I think we're going to need something where, if a translation goes to 0, it dies. That'll allow you to vote against your bone-headed suggestion and have it kill it.

>12 timspalding:
Yes, that should happen. Even if, for whatever reason, you can't kill a translation at 0 (even I know that computers are stupid sometimes), at least have a net vote of -1 be a tag translation's death warrant.

So are Wikipedias going to be tapped that haven't already been tapped? Going on epo.LibraryThing.com doesn't show tag translations from eo.Wikipedia.org.

But apparently adding a translation to fantasy does know that it's the translation used on Wikipedia.

When, on a tag page, I attempt to enter an alternative translation for a language already in the list, I get the following error:

Duplicate entry 'spa-5334-9568035' for key 2
INSERT INTO tag_translations ( tagt_ttkey_from, tagt_ttkey_to, tagt_language, tagt_usernum, tagt_stamp ) values ( 5334, 9568035, 'spa', 131513, 1339551403 )

- fatal error (31.1)

For example, on the page for the tag world literature, the Spanish translation "Anexo:Literature por idioma" is presented for voting. I believe a better translation would be "literatura mundial" ("literatura por idioma" means, literally, "literature by language" and therefore corresponds more closely to "foreign language literature"). When I try to enter "literatura mundial" as an alternative Spanish translation (using the "Add tag translation" link), I get the error above.

(Edited to Fix Link)

Also, the list of languages in the "Add tag translation" dropbox is not sorted.

http://epo.librarything.com/tagcloud.php isn't translated; this seems to be a very useful tag feature to translate, since it gives a quick list of useful tags to translate.

Do we have to be logged in to the other language site to make proposals?

#18: No; just click see all, add tag translation, and remember to set the language correctly.

Thanks.

For some reason, the SE code on Wikipedia is for Sámi, not Swedish. Swedish is SV. There are some interesting tag translations for Swedish now up here, I can tell you... ;)

#21: The se code on Wikipedia is for Sami instead of Swedish because the language tagging standard (ISO 639) says that se is for Sami. Country codes are in a separate name space then language codes.

>18 Nicole_VanK:, 19

You must be on a non-English site. The tag translation option is not available on librarything.com.

Yes it is. It's not available for setting translations into English though - which I think is a bit odd.

Not that I mind, I'll do my bit. But this is going to be one humungous task.

> 24

Where then? The option to "add tag translation" does not appear when one follows your advice in #19 to "click see all".

>7 timspalding:

Might just be Swedish? Or are you not supposed to be done with shifting to svwiki?

When, on a tag page, I attempt to enter an alternative translation for a language already in the list, I get the following error

Duplicates only occur when the same base tag ("love") acquires the same translation ("amor") in the same language ("spanish"). I've stopped the error from showing, but is this what you were trying to do?

Do we have to be logged in to the other language site to make proposals?

You have to be logged into whatever site you're on, but you can propose translations on any language from any site.

http://epo.librarything.com/tagcloud.php isn't translated; this seems to be a very useful tag feature to translate, since it gives a quick list of useful tags to translate.

Yes, I agree. That's one issue. The other is the Swedish issue.

You must be on a non-English site. The tag translation option is not available on librarything.com.

No, it's available on all the sites.

Where then? The option to "add tag translation" does not appear when one follows your advice in #19 to "click see all".

I see it. Perhaps you're not signed in?

> 26: It's way down on the tag page. Even below "Google Books Ngram Viewer". First you see "see all", click that and you'll also get an "add tag translation" option.

> 29

Oh!

That's odd. If I'm on one of the other language sites, I get to it from the tag list on the main book page, which is what I thought you were referring to. The option doesn't exist there on the .com site that I can see.

> 28

I see it. Perhaps you're not signed in?

I'm always signed in! But see above. It seems to be a matter of having the option on the tag page, but not on the main book page (unlike the non-English sites).

I'm always signed in! But see above. It seems to be a matter of having the option on the tag page, but not on the main book page (unlike the non-English sites).

Ah, but the "option" on book pages is just to show or not show, not to edit translations for a specific tag.

> 31

But I can translate the tags from there on the non-English sites, just by clicking on the tag, and that's where the confusion arose.

I gotta get new glasses... I read the subject line as "Tao Translations" and thought someone was looking for recommendations....

Tim, if you made the translated tags show up in the all LT tag cloud, that might help us knock out some highly used (and therefore high priority?) tags.

In other news, I'm having fun learning how to say things like "wormhole" and "language death" in French.

I'm very sorry it's "trou de ver." I mean, I now know what "worm" is in French, but they should have their own cool name.

Speaking of ver, however, maybe this is what Prinn's De Vermis Mysteriis is all about.

I wouldn't know, I haven't read it. I am comforted to hear you haven't either.

>I'm very sorry it's "trou de ver." I mean, I now know what "worm" is in French, but they should have their own cool name.

Eh, I don't mind calques. They certainly make this kind of work easier.

I must be completely stupid but how exactly do you suggest a translation?

When you translate tags, should you capitalize or not capitalize common nouns? Common noun translations from Wikipedia have their first word capitalized. With member translations it varies.

Where can you alert and discuss bad translations?

I must be completely stupid but how exactly do you suggest a translation?

At the bottom of the tag page there's a section for tag translation. Depending on whether you're on English or another language, the link may be there or beneath a "see all" link.

Any timetable for when you think you have fixed Swedish -- including deleting all the prepopulated translations from the Sámi Wikipedia?

The language list when you add a translation is now translated and sorted.

More features: http://www.librarything.com/topic/138516

Forgive me if this has been discussed elsewhere...

Should tag translation proceed like tag combination -- that is, should we only use words that have no other meanings? For example, the Spanish translation for periodical is periódico, which is also the Spanish word for newspaper. Niños means both children and boys (unless you use muchachos, which also means both boys and children).

At any rate, if we were to translate both periodical and newspaper as periódico, and both boys and children as niños, would this lead to some sort of emergent tag ambiguity, where at some deeper level periodical is equated with newspaper and boys is equated with children?

I don't think so. The aim, as I understand it, is to have users of the other language sites get to view all tags in their own language.

But I agree the fact that there will sometimes be multiple versions of the "same" tag will cause some confusion on the other language sites.

I do try to keep things disambiguous though. For example: I think Dutch for "American" should be "Amerikaan(s)", since Dutch uses different forms for the noun and the adjective. Not elegant, but I think that's the best we can do.

(edited to fix typo)

But some tags provide real problems. For example "Stamps" - it's mostly about postage stamps (apparently), which would translate to "postzegels" in Dutch. But it's sometimes about rubber stamps, which would translate to "stempels" in Dutch.

Translating it to "Postzegels / Stempels" would seem ridiculous to Dutch speaking people. Since the words are totally unrelated, in our mind there is no association between the objects they represent.

More problematic translations, this time in Finnish:

Book club was translated as kirjakerho. The translation is correct, if it is a book sales club. If it is a book discussion club, it should be translated as lukupiiri. It is often difficult to know, which kind of club the tagger was thinking about.

Which is the best way to translate the tag? The Finnish tag lukupiiri is far more popular than kirjakerho. Is this a good way to estimate, which meaning is used more often?

>46 Nicole_VanK: Yeah, I keep coming across examples like this too -- most of the time, I choose the more likely translation.

Just now, however, I submitted two Spanish translations for borders -- fronteras (e.g., borders between countries) as well as bordes (i.e., edges). Looking at the books that have been tagged borders, it seems that this label has also been used quite a bit as a synonym for boundaries. Would it be overkill to add a third translation (límites), since voting will result in only one translation being selected overall?

Which is the best way to translate the tag? The Finnish tag lukupiiri is far more popular than kirjakerho. Is this a good way to estimate, which meaning is used more often?

I suspect so. You can't be sure. Maybe in very difficult cases the best translation would be "A or B"?

Some of the auto-translations are very funny. Apparently "Ian Hodgkinson" is French for vampire ;-)

It seems that the Northern Sami "translations" are still there for Swedish. Is there any way to get a list of them?

Any chance of Tim (or any other of the PTB) chiming in on this predicament?

http://www.librarything.com/topic/138714

Please.

Sorry. I'll attend to it tonight. Apologies.

No worries.

Sorry. Conference is killing me. Tomorrow/today after it ends and Abby, Kate and I are at airport.

Relax.

Sure we want an answer - if we can get one, but we need you guys alive and well too.

Since you seem to be back on your feet Tim, please, if you get a chance: http://www.librarything.com/topic/138714