[REL] Nena (1997)

User avatar
Night457
Global Moderator
Posts: 5334
Likes:
Joined: Sat Dec 28, 2019 3:44 pm

Re: [REL] Nena (1997)

Post by Night457 »   0 likes

Ooh, you found the Spanish subs? Where? Can you post the original Spanish as an attachment? What translator did you use? Subtitle Edit / Google-translate? Did you try DeepL instead, which may be better?

https://www.deepl.com/translator

If no one can do a better translation I will happily take a crappy one.
mexlink
Posts: 148
Likes:
Joined: Fri Jan 14, 2022 12:59 pm

Re: [REL] Nena (1997)

Post by mexlink »   0 likes

Night457 wrote:Ooh, you found the Spanish subs? Where? Can you post the original Spanish as an attachment? What translator did you use? Subtitle Edit / Google-translate? Did you try DeepL instead, which may be better?

https://www.deepl.com/translator

If no one can do a better translation I will happily take a crappy one.
I can confirm deepl is MUCH better than other online translation tools. I used it a couple times to translate mangas.
User avatar
Night457
Global Moderator
Posts: 5334
Likes:
Joined: Sat Dec 28, 2019 3:44 pm

Re: [REL] Nena (1997)

Post by Night457 »   1 likes

mexlink wrote:I can confirm deepl is MUCH better than other online translation tools.
I have seen that claim from a number of people here, but I do not believe it. I have done side-by-side comparisons of SubtitleEdit (Google Translate) and DeepL translations of the same document. Both translators have their merits, and both produce some errors. The DeepL errors seem to have the "benefit" of being obvious and glaring: many words are left as-is in the original language, and some of the translations are just plain stupid. Google also produces stupid translations; I just do not see the evidence of the superiority of one over the other.

What I find most irritating about DeepL is the necessity to convert a plain text (.srt) file to .docx in order for it to translate, then I have to convert it back. Heck, pretty much the only documents I am going to translate are subtitles!

How much a DeepL translation may be better than others I suppose depends on the language. It may be better with some than with others. Whenever I do serious subtitle translation, I am comparing multiple translations. If I just want something quick and easy, it will be SubtitleEdit.
I used it a couple times to translate mangas.
DeepL is good for Japanese? Awesome! I will keep that in mind.
lildrm
Posts: 43
Likes:
Joined: Thu Feb 11, 2021 12:22 pm

Re: [REL] Nena (1997)

Post by lildrm »   0 likes

Night457 wrote:Both translators have their merits, and both produce some errors.
True. I love Deepl for translations of longer texts, but I was unable to achieve the same quality when translating subtitles. Often getting better results from Google.
User avatar
Night457
Global Moderator
Posts: 5334
Likes:
Joined: Sat Dec 28, 2019 3:44 pm

Re: [REL] Nena (1997)

Post by Night457 »   0 likes

I wonder if the fact that a subtitle file is half numbers and half text confuses the translator? SubtitleEdit maybe avoids that by loading the .srt file and just sending the text part to Google Translate.
User avatar
DreamScape
Posts: 296
Likes:
Joined: Tue Sep 15, 2020 7:06 am

Re: [REL] Nena (1997)

Post by DreamScape »   2 likes

lildrm, I think Night457 really nailed the problem with Deepl, which might also be why you find it subpar for short phrases.

Consider the Swedish phrase "van fan"

Deepl seems totally confused by it

[Image]

Meanwhile Google understands

[Image]

But if we fix punctuation Deepl will get it

[Image]

Deepl seems to be very sensitive at times with surrounding context, which can lead to words left untranslated in subtitles and such if the phrase is too short, has incorrect or missing punctuation, or is split across multiple lines.
User avatar
Night457
Global Moderator
Posts: 5334
Likes:
Joined: Sat Dec 28, 2019 3:44 pm

Re: [REL] Nena (1997)

Post by Night457 »   1 likes

DreamScape wrote:lildrm, I think Night457 really nailed the problem with Deepl
Wow, that would be like the first time EVER! I was just guessing.

Thanks for your analysis with examples. It makes it much clearer why DeepL works better with longer texts, as lildrm noticed.
Post Reply