The Irreplaceable Art of Translation | Hacker News Reader

The Irreplaceable Art of Translation

prospectmagazine.co.uk

40 pointsVigier21 comments

For the indie computer games I make I work with a number of localisation freelancers and it's fantastic to see the creativity of people who love games and love language. My current game[0] for instance is a game with lots of made up monsters in it, that have their own made-up English names - and it's often the appropriate thing to do to create different monster names per-language that fit with their designs.

Obviously necessary for Korean/Japanese/other non-latin alphabet languages, but my Spanish localiser does a fantastic job in this too. For example: Hellion -> Infernodrilo, Grondomut -> Cienomello, Gastroquin -> Explosapo. No, they're not direct translations (if that were even possible) but they fit with the visual design of the creatures.

There's a lot of trust you have to have with the people who make sure a creative product retains meaning and expression in another region and language, but also is adapted where it needs to. It's also a lot of fun to see something I made enjoyed by people I wouldn't ordinarily be able to communicate with. Irreplaceable indeed!

[0] - https://store.steampowered.com/app/654960/The_Eldritch_Zooke...

> My current game[0] for instance is a game with lots of made up monsters in it, that have their own made-up English names

"Hellion" doesn't really fit that theme.

Why not?

It is not a made-up name; it is an ordinary English word. It's not particularly monstrous either - you'd typically use it to describe a child prone to misbehaving.

The Spanish translation given looks like "hell crocodile", which is radically different than the mischievous little kid suggested by "hellion".

"Hellion" in a fantasy setting has a different connotation than the meaning you refer to. They're represented as hellish fiery creatures in a lot of fantasy games.

I hope I'll never have to read a novel translated by "AI". It's already bad enough to see auto-translated documentation for Microsoft APIs and software by default, recently even auto-translated localization strings in Windows itself, or having to suffer from badly auto-translated item listings on ebay and other websites. Please stop pretending that automatic translation is a silver bullet that can stand on its own.

Meanwhile I can't wait for the realtime voice/phone translator, which would be much more useful than me spending years learning a new language properly.

And I can't wait for flying cars.

Except it's not a pipe dream. IBM and I think Google showed something like that a few years ago.

And looking at it now, Skype has something: https://www.skype.com/en/features/skype-translator/

And Qualcomm is trying to integrate it with the hardware, which should reduce lag to an acceptable level:

https://www.qualcomm.com/news/onq/2020/03/26/how-snapdragon-...

Privacy advocates will have a field day with this stuff, though.

I'm not too worried about the lag, moreso about the quality of translation. It will likely be useful for basic conversations where the participants simplify their speech, which often is all you need. Or to get a rough idea of a more complex conversation.

But it's far from obvious that we'll be able to go from there to anything comparable to the level of conversation you can have after spending years learning a new language properly.

Perhaps nuclear fusion is a better comparison than flying cars (which arguably would be a bad idea even if we had the tech).

Deepl is already so good people don't realize they're taking to someone with zero knowledge of their language.

Shouldn't be that hard to create a voice-to-voice translation engine.

Specifically for Europe, the idioms and jokes are the 20%, we're already very similar to each other even though many don't want to believe it.

Flying cars aren't exactly a pipe dream either. If we wanted them half as much as we want normal cars, we would have them by now. The right group of rich people could make them happen too.

No two translators would find the same words. There is nothing like a correct translation. Language is loaded with meaning beyond words and written by authors living their lives in societies and across the ages.

> "There is nothing like a correct translation."

Perhaps not but, having read a lot of translated material over the years, there are plenty of remarkably awful translations.

Maybe there's some threshold that computers cannot cross. But the constant improvements show no sign of stopping soon. I wouldn't bet against it, given a long enough time period.

Technology tends to follow a sigmoid curve. First nothing happens, and then it seems like the world is changing and it is unstoppable, and then it falls into a new normal, not changing much. We never know when it slows down, reaching the third stage, until it's happened. For all we know, this is the furthest we get. Personally, I don't believe that, but food for thought.

"But the constant improvements show no sign of stopping soon."

True. A GPT-3 type approach probably has a big enough window and enough data to pick up on idioms.

But we have a long way to go with the Asian languages. All the European languages seem to machine-translate to each other reasonably well at this point. Chinese to English is still not very good.

"No way can a machine replace me", said everyone ever replaced by a machine.

Accurate translation is a rather simple matter imo, you "just" need a big enough database (built by humans, heh).

I also prefer reading the original jokes and idioms and looking up their meaning myself. It's more fun.

Large enough dataset + some way to see context. There could be an ambiguous word that is only found through accompanying media. A multimodal AI would be needed for that task. Maybe something like https://blog.google/products/search/introducing-mum/ would be up for the task, maybe not. But you are correct, people say a machine can't replace them until they do.

Sounds like what Deepl is doing. Surprisingly accurate between some languages, and the translation changes depending on context. But it needs way more human input/corrections before it can be near perfect.

I was reading an article talking about book translation and translation styles a couple of days ago but I can't seem to find it again. In the meantime, have a link to some quotes on localization and translation: https://legendsoflocalization.com/literary-translators-comme...

Reply on news.ycombinator.com