On the road to... Pillars of Eternity 3

[Madscientist]

44 minutes ago, wih said:

This theory will have to be expanded probably, because at least Josh Sawyer has some clue : https://jesawyer.tumblr.com/post/169284230541/the-problem-with-tokens

Being aware of the problem is the first step, but solving it is something different.

In PoE1+2 I have the feeling that the people who made the translation were missing context in several cases and the people who put the text into the game did not check what they got from the translators. Only when the translators can directly talk with the devs the result can become good. RPGs have more complex settings and rules than most other games and this stuff is often hard to understand in the original language. It is almost impossible to translate something when you have problems to understand the original.

I guess every developer in the world speaks english. Among other reasons, almost all programming languages are based on english. But the problem is to be a good writer. Knowing a language is not enough to write good texts in this language. And even if you can write well in one language it does not automatically mean you can write good stories in another one, even if you speak several languages. Maybe european companies are more international than american ones? For example Larian is in belgium, so I guess most employees know english, dutch and french and a lot of them know german too. Owlcat has also many employees who know more languages than just russian and english.

To be fair, the trails series may be a bad example. They have great translations, but they are translations made for a game that is already finished. So the translators could play the full game in japanese when doing the translation and they could ask the devs if this and that is meant in this or that way. It is something different when you translate stuff when the game is still in developement.

At least for germany I can say almost everyone learns english at school and most people learn a second foreign language as well. For me it was french, but I have forgotten everything since then. I know several native english speakers and some complain that its hard to learn german in germany because people automatically start speaking english when they hear that your german is not so good. I do not know much about the US school system, but given the history and the other countries around it would make sense if many people learn spanish and maybe french.

PS: I know that Josh knows some basic german, some of his posts are in german. But he admitted it is not good enough to translate complex texts in a very good way so he cannot check the translation. I do not know if he knows other languages.

[xzar_monty]

4 minutes ago, Madscientist said:

But the problem is to be a good writer. Knowing a language is not enough to write good texts in this language. And even if you can write well in one language it does not automatically mean you can write good stories in another one, even if you speak several languages.

All of these claims are true. What I would add to the question of translation is that being a good translator is a different skill altogether, i.e. different from being able to speak a language very well and different from being able to write very well.

To take an analogy from music: playing an instrument, being able to improvise on it and being able to compose on it are three entirely different skills. There is some overlap, but even that is not guaranteed: you can be a superb classical musician but still not able to compose or improvise at all.

Likewise, knowing a language, being able to write well and being able to translate from one language to another are three entirely different skills. Apparently, what Obsidian lacked was 1) good translators and 2) quality control to point out that the translations they received were rubbish.

[Madscientist]

17 minutes ago, xzar_monty said:

All of these claims are true. What I would add to the question of translation is that being a good translator is a different skill altogether, i.e. different from being able to speak a language very well and different from being able to write very well.

To take an analogy from music: playing an instrument, being able to improvise on it and being able to compose on it are three entirely different skills. There is some overlap, but even that is not guaranteed: you can be a superb classical musician but still not able to compose or improvise at all.

Likewise, knowing a language, being able to write well and being able to translate from one language to another are three entirely different skills. Apparently, what Obsidian lacked was 1) good translators and 2) quality control to point out that the translations they received were rubbish.

You are right and I do not know the answer.

Another thing is the kind of text to use. I have written and translated stuff like manuals or other technical texts in german and english, but I could not write an interesting fantasy story. A technical text needs to be clear and easy to understand and you should avoid any emotional stuff in it.

[wih]

1 hour ago, xzar_monty said:

What he's writing there does not demonstrate having a clue, although it is not proof of cluelessness, either. What he's talking about is something one obviously has to know if one's going to get involved in something like this. The very basics, meat and potatoes. Does Josh Sawyer actually speak a language other than English? Not that he has to, of course.

You will have to keep in mind that Obsidian's translations (most of them, at least) are demonstratably awful, so Sawyer's understanding of this particular phenomenon is not worth much.

I think Sawyer studies German. I was suggesting that the cause of those bad translations is maybe something else than simply cluelessness.

[xzar_monty]

1 hour ago, wih said:

I think Sawyer studies German. I was suggesting that the cause of those bad translations is maybe something else than simply cluelessness.

Cluelessness may be putting it strongly, I can agree with that, but something somewhere went very wrong, and apparently Obsidian was unaware of it (because the horrible translations went ahead), and that in itself would qualify as some kind of cluelessness.

[Boeroer]

I do not think that Josh checks translations at all since his job is a different one (too expensive to waste time with checking localisations) and he and Obsidian have to trust the translators to do a proper job.

He is fluent in German (and knows Spanish and some French iirc) - but I guess it's not enough to determine if a translation is really good or just understandable. 

I'm fluent in English but I wouldn't want to/couldn't determine if a translation from German to English is actually really good or just not obviously bad.  

[xzar_monty]

9 hours ago, Boeroer said:

I'm fluent in English but I wouldn't want to/couldn't determine if a translation from German to English is actually really good or just not obviously bad.  

This is fair enough, and reasonable. Having worked in the field for a couple of decades, and having dealt with translations of highest-quality literature, I can assure that what you're talking about is not easy, unless of course a translation is just awful. English is not my first language, but I know it well enough to make good judgements there. In my other foreign languages, no, although I have translated a little bit of poetry from French.

Edited July 26, 2020 by xzar_monty

[Wormerine]

I am also afraid that PoE aren’t the easiest to translate. It’s use of language is unusually sophisticated for a video game. For Tolkien it took 2 tries to translate it properly into Polish, and honestly the strength of 2nd translation is making changes to the source material. I have been wondering for a long time, if Bioware basic and simple language was fully intentional for world-wide appeal. 

[thelee]

On 7/25/2020 at 7:08 AM, Madscientist said:

In PoE1+2 I have the feeling that the people who made the translation were missing context in several cases and the people who put the text into the game did not check what they got from the translators. Only when the translators can directly talk with the devs the result can become good. RPGs have more complex settings and rules than most other games and this stuff is often hard to understand in the original language. It is almost impossible to translate something when you have problems to understand the original.

i would strongly bet that obsidian just outsourced it to a third party and the translations were done by some thankless workers who couldn't circle back with the designers. because it's cheap.

i work at a major web company and internationalization is part of my job. we do string tokenization the way sawyer does. we also have to provide context on each sentence or string to give the translators some remote clue what on earth we are talking about, because a sentence in abstract could mean a lot of things. even then, i get reached out to by the translators themselves (they have that authority and capability to do so) to get even further clarification on where a piece of text might be seen, and what other text might be nearby, to further get a better translation. and our translations are merely adequate (apparently we are known to be weak in some cultures where there is a gap between the formal language and every day use of the language, especially if a pidgin english is part of it, because our translators focus on getting the formal language correct isntead of what is actually spoken), and our text is pretty simple. there's no complex lore to encapsulate or prose to handle, just "this button does this." make the text more complex and go for a cheaper interantionalization business structure and i could easily see the translation going off the rails bad.

 

kotaku did a huge series on FF7's original japanese -> english translation, and i suspect that's pretty much how obsidian ended up. hopefully now that they have microsoft resources future internationalization work can be much better.

Edited July 28, 2020 by thelee

[xzar_monty]

9 hours ago, Wormerine said:

For Tolkien it took 2 tries to translate it properly into Polish, and honestly the strength of 2nd translation is making changes to the source material.

Making changes to the source material, if it means changing what Tolkien wrote, sounds unbelievably dubious (not to mention directly against all tenets of proper translations) and is something the Tolkien estate would not accept.

I cannot comment on the Polish versions of Tolkien, but it's fairly well known, for example, that the first Swedish translation of LotR (which I also have not read) is poor. This happens, as unfortunate as it is.

[xzar_monty]

2 hours ago, thelee said:

our text is pretty simple. there's no complex lore to encapsulate or prose to handle, just "this button does this." make the text more complex and go for a cheaper interantionalization business structure and i could easily see the translation going off the rails bad.

Yep. Deadfire is ambitious both in style and in content. If Obsidian simply outsourced the translation to a third party, as you suggest, it tells me that the company either did not understand what it was doing or it did not care. The second possibility sounds unlikely, so I would bet that it just didn't understand what it was doing, which comes back to my earlier point about Obsidian being clueless.

Your description of the difficulties of translation, even of simple texts, sounds authentic and apt.

[Wormerine]

54 minutes ago, xzar_monty said:

Making changes to the source material, if it means changing what Tolkien wrote, sounds unbelievably dubious (not to mention directly against all tenets of proper translations) and is something the Tolkien estate would not accept.

Mostly refering to things like names were changed to be intuitively convey what author wanted to convey. Poetry modified to still somewhat sound good and convey the meaning. That kind of stuff - that wasn't Shrek kind of translation, where jokes are completely different. Tolkien uses language in very specific way and that's something that can't be ignored nor easily translated.

In PoE specifically names could be problematic (Dyrwood has a meaning and will take some thinking to properly translate, or most likely find fitting substitute). Also different races speak in different manners. How will one translate Huana's speach pattern? Those are things that need to be taken into account and 1) Indeed require context which translator might/might not have 2) will take time to think through and review, and I doubt deadlines were generous. And I think that if such details are lost PoEs loose a lot of it's charm. Perhaps that is a reason why I didn't like Obsidian games until I lived for couple years in English speaking country and became more fluent with the language. 

[Boeroer]

Translation was indeed outsourced.

 

[xzar_monty]

9 hours ago, Wormerine said:

Poetry modified to still somewhat sound good and convey the meaning. [...] Tolkien uses language in very specific way and that's something that can't be ignored nor easily translated.

This is very hard to understand -- but then it's also off topic, so maybe we should just stop. I mean, it's impossible to understand what "poetry modified" might mean without seeing any examples. Also, most authors use language in a very specific way -- Tolkien's use of English is quaint and erudite but not especially problematic to translate. James Joyce's English is problematic to translate, for example.

[Sven_]

https://wccftech.com/third-pillars-of-eternity-game-happen-excited/

A third entry in the Pillars Of Eternity may happen in the future, but only when the developers will be excited for it.

[Wormerine]

That’s all I wanted to hear. I shall be waiting.

[J.E. Sawyer]

I have never heard of any game developer doing their localization in-house, not in 1999, not in 2020.  Every game I've worked on and every company I've ever talked to always outsources their localization.  Localization companies specialize in it.  There's no reason for even a mid-sized game dev studio to keep a loc staff on hand unless it's a single person doing coordination.  And that single person is usually a production role, not a polyglot translator.

Even localization companies often outsource one or more languages that their normal staff can't cover.  Deadfire shipped in 10 languages: English, French, Italian, German, Spanish, Portuguese, Polish, Russian, Korean (separate SKU), and Simplified Chinese.  I have never seen an entirely in-house localization team that covers all of those languages.  If they exist, I assume their per-word translation costs are phenomenally high.

We do have fluent and native speakers at the studio occasionally play the game to spot check how things are going, but there are hundreds of thousands of words in our games.  Even a native speaker working as a game developer does not have time (especially late in the project, when the loc'd versions are close to final) to play through the game entirely in their native language and write up the bugs that they find.  I played the game for maybe 5 or 6 hours in German and caught a bunch of bugs that I sent back to the loc team.  I also played the game for a couple of hours in French and caught a few bugs (my French is not great but the bugs were really obvious).  That's a drop in the bucket, content-wise.  And I'm fluent in German, not native.  Something that macht Sinn to me might actually klingt falsch to a native speaker.

Our audiences have always expected translations to ship "day-and-date" with the English versions.  With this volume of text and this number of languages, I have yet to see a time- and cost-feasible solution for not only translating all of the text but comprehensively reviewing it in situ and sending it back for per-string spot-fixing before launch.  I know people were disappointed by the German and Italian (at least) translations, but it's not like our company is doing something uniquely weird and stupid.

[Sven_]

15 hours ago, Wormerine said:

That’s all I wanted to hear. I shall be waiting.

Well, that's basically a confirmation of that it's not dead -- yet. Still far from a confirmation of an actual game ever happening, mind.  

In the meantime, there'll be Black Geyser, which looks suspiciously Pillars-like. Will be interesting how well Wrath Of The Righteous does too. If the latter becomes a hit rather than being hit by the "sequels sell worse syndrome", that may trigger the business side of things at Obs too.

Currently I personally view this as slightly higher a chance as the chance of Alien:Isolation 2 ever happening. The first was equally clearly a passion project with many fans of the IP involved, and they've been toying with the idea of a sequel on the occasion, but the game just about broke even sales-wise (and didn't meet Sega's expectations). At least the team hasn't been disbanded, unlike Isolation's.  Chances might be higher if the option was assessed of scaling things back a bit where possible (smaller project on a smaller budget) -- again, none of the BIS-developed Infinity Engine games ever broke sales records as far as we know and still made their money.

 

 

Edited August 6, 2020 by Sven_

[xzar_monty]

17 hours ago, J.E. Sawyer said:

With this volume of text and this number of languages, I have yet to see a time- and cost-feasible solution for not only translating all of the text but comprehensively reviewing it in situ and sending it back for per-string spot-fixing before launch.  I know people were disappointed by the German and Italian (at least) translations, but it's not like our company is doing something uniquely weird and stupid.

There is no such solution. The writing in Deadfire is ambitious, literary and challenging, and while its quality doesn't approach good literature, it is definitely not pulp, either. There is verve and poise in the choice of words, and while it occasionally gets too verbose for its own good, it is generally enjoyable and engaging.

You are quite right that Obsidian is not doing something uniquely weird and stupid, but this in itself does not say much. I have over twenty years' worth of experience in this field and close to eighty books to my name, and while I'm not an authority as such, I know fairly well what I'm talking about. The problems you describe (in terms of the amount of text in the game and the amount of time available) are real, and they are exacerbated by the fact that the text is ambitious, i.e. pulp would be a lot easier to translate. But the solutions chosen are far from ideal: essentially, the translators would need a lot more time, which in turn would cost a lot more money, and I can see how this might not sound good to Obsidian. But the fact remains: given the time and resources allocated to the translation project, the results can be quite poor, and while pointing this out may feel harsh, it is justified.

Edited August 6, 2020 by xzar_monty

[thelee]

I know this is not a good general solution, but do there exist localization mods for deadfire?

[Boeroer]

I think @Phenomenum did an improved Russian mod or something?

There were also improved German mods for PoE back in the days. They discussed German phrases and names in the PoE modding subforum.

I think to most of the players who really need a German version (because their English is too bad) the localisation actually sounds alrighty...

If you'd ask me how I would handle localisation of a text-heavy game: I'd provide an easy localisation framework (as part of a modding framework for example) and let the players do the localisations. Maybe make it a challenge with a badge. The ultimate translation! Would be cheaper I guess and would result in better quality. But you couldn't have localisations at release then.

 

 

Edited August 6, 2020 by Boeroer

[Boeroer]

@J.E. Sawyer 

Hey! Welcome back, welcome back!

 

[xzar_monty]

7 hours ago, Boeroer said:

If you'd ask me how I would handle localisation of a text-heavy game: I'd provide an easy localisation framework (as part of a modding framework for example) and let the players do the localisations. Maybe make it a challenge with a badge. The ultimate translation! Would be cheaper I guess and would result in better quality. But you couldn't have localisations at release then.

After writing my reply to Josh Sawyer, I started wondering about the same question. It would be interesting to know the finances of, say, the Italian translation. I.e. how much it cost and how much business it generated among people who only play in Italian. Of course I'm not going to get that information.

If I was Obsidian, I would quite possibly just ditch the idea of localization, but this is mainly because I know how difficult it is to get a good translation of something like this, and because I certainly wouldn't accept a bad one. But I am not Obsidian, so of course I cannot really say. There are many questions Obsidian has to consider when making decisions like this, and I know nothing about all the relevant things.

It has to be remembered that there are also bad translators, like there are bad writers and bad singers. Who work in the field despite their obvious shortcomings. So at least some of the fault may lie there, too. For example, it is fairly well known among translators that the first Swedish translation of Lord of the Rings is poor. Tolkien writes about this in his letters. (I have not read it, so I have no personal comment to make.)

By the way, your suggestion is potentially a good one, but it would likely only appeal to amateurs. Some of whom may be excellent, don't get me wrong. But as a professional in this field, I would never take part in a translation job like this for free.

[wih]

So in the end, localization of text heavy games is an unsolved problem currently. Such is life, I guess. Obsidian also chose to release the game despite knowing for sure there were still many bugs left. But what can you do? Taking five more years to make the game perfect isn't ideal either.

Edited August 7, 2020 by wih

[Boeroer]

@xzar_monty: Well, amateurs or not - if players like Phenomenum say that the Russian localisation is not acceptable and start do do their own then the outcome that those amateurs would provide can't be worse than the original. It might not be top notch literature either, but better than what's out atm. AND Obsidian weren't to blame. Win-win.

The German mod for PoE (at least the parts I read about in these forums) was def. way better than the original. But having said that: those dedicated guys were able to work with an already existing translation and didn't have to do one from scratch - which most likey is more difficult and time consuming(?). 

It's also unknown (at least to me) if there would be big complaints because Obsidian didn't ship a game with localisations like they used to.

Edited August 7, 2020 by Boeroer