Contributing with google translate/Deepl

General Discussion
41

I would also think that viewers would be more inclined to grade a bad translation than a good one, creating a skewed grading system.

4 Likes
42

The viewers can only judge if the subtitles sound natural in their own language or wooden/idiotic/wrong to gobbledigook and everything in between (I have seen 99% gobbledigook subtitles in my language here on Viki, I kid you not). Of course, viewers may not be able to judge how accurate your translation is. I’m just suggesting some kind of reward for those translators who don’t mindlessly google translate.

1 Like
43

At this point any actual system would be an improvement :smiling_face_with_tear:

2 Likes
44

Actually, even in their own language, not everyone can do that.

That would be great, but still very hard to implement in an effective way.

2 Likes
45

Since I’m a viewer (reader), I think I can tell the difference quite well now :joy: Maybe the viewer should grade the individual translator
:thinking: wonder if Viki doesn’t have so many translators anymore?! :joy: escaped for the sake of honesty
 Sometimes it’s really bad


2 Likes
46

I’ve seen people defending machine translations here and there. I don’t think machine translations are ever good enough for subtitles. Here’s why.

  1. Machines don’t see the context of the scene, “they” don’t know what has gone on before, what is the relationship between the speakers. They also don’t know the characters, their motivations and fears. They don’t know the nuances. Fun fact: Machine translations can’t consistently use formal/informal language as English lacks this distinction. English also lacks distinction between singular and plural “you”. On the other hand, English has gendered pronouns, while some languages don’t. These 3 things already pose a siginificant confusion risk.
  2. MT uses bland language. Real language use by real people is full of idioms; MT is very much idiom-free. There is no flavor. This causes unnatural, wooden, too official, stiff language. Sometimes unintentionally funny language. Also, MT can’t really do anything with non-contemporary language, for example, expressions and words you need to use in historical/costume dramas.
  3. Sometimes MT just gives you gobbledigook. It can make mistakes.
  4. Robots don’t have your experience, knowledge, instinct and judgement. That’s the truth. They are not moved by the story and the talent of the writers and are not motivated to render some beautiful/moving/funny lines into your language to create the same impact. You are the heart and soul through which all this is filtered. Machines don’t have hearts and souls. Yet. :wink:
  5. Literal translations are not always the best translations. This is where your creativity comes in.
  6. Sometimes MT gives you convoluted sentences. For a subtitle to work, they should capture the full meaning of the original sentence in a way that can be instantly grasped by the viewer, without having to pause the scene and read the sentence thoroughly in an attempt to wrap their heads around it.
    For these reasons, MT is not even a good foundation for subtitles. You’ll find that even if you use it, correcting it amounts to starting everything from scratch. My view is that for good subtitles, you need to understand what is going on in a scene and then to translate the scene rather than decontextualized, individual sentences in it, if that makes sense.
    MT makes the viewing experience sadly awful, although many viewers may feel it is better than nothing. But do we really need subtitlers to just add automatic translations? Is there any added value? I think some people will argue that it is speedy. There is that, sure.
11 Likes
47

Sadly, that is the case. Bad subbers or machines
 as long as they produce subtitles FAST, the majority of Viki’s source of income does prefer that above quality that they have to wait for.
Personally, I can get so irritated about bad subtitles that I tend to switch to other languages if I’m watching something, but I am not exactly the majority here


8 Likes
48

@mirjam_465
That’s because you are so wonderfully versatile in languages. You can simply :smiling_face:

In fact, I’ve now also noticed that the machines translate really badly when they are used. :rofl: even though I use the machine myself


I also switch to English so that I can somehow translate it for myself
 :relieved: somehow :sweat_smile:

1 Like
49

Hah! Have you ever watched K drama on other streaming sites? It irritates me so much on H***U, because if you choose subtitles, you received close captions with very inappropriate descriptions of sounds such as when men are laughing the caption is “giggling” and whenever footsteps are heard, even for a lady wearing spiked heels, it’s “heavy footsteps”. Have you ever heard paper “clattering”? Do the hearing challenged even know what “quirky music” is? Even worse, some dramas are only available dubbed. If you like the sound of Seo Ji Seob’s voice, too bad. The dubbing sounds like it is from a Gungfu movie from Hong Kong.

6 Likes
50

LOL :rofl::rofl::rofl::rofl::rofl::rofl::rofl:

You just made my day with this funny comment!! It’s definitely what you wrote! And what about the websites that, when is playing a soft music, they put “high-pitched whistle” in the subtitle to indicate that there’s music background!!

KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK = LOL LOL LOL LOL LOL LOL LOL LOL LOL!!!

2 Likes
51

A recent report on MT and Generative AI translation identified the following translation mistakes:

Translation mistake typology

In this analysis, we identify the following types of translation errors:

  • Mistranslation
  • Literal translation — a near word-for-word translation that lacks necessary adaptation
  • Translated DNT(do-not-translate) — a part of text that should not have been translated (like an acronym or proper noun) was translated
  • False DNT —a part of text was mistakenly left untranslated
  • Mishandled DNT — DNT was handled inconsistently or incorrectly
  • Omission
  • Untranslated —the entire text or a significant portion was left in the original language
  • Hallucination — a GenAI-specific issue where the output has no relation to the original text
  • Punctuation
  • Grammar
6 Likes
52

Most of these issues are logical limitations of MT, but I get a chuckle out of the fact that machines hallucinate too.

7 Likes