STT Speech To Text acceptable?

[0pener]

Nowadays automatic text to speech is really much better than a decade ago, for example whisper, vosk, kaldi etc... Everybody knows youtube's auto subs, eh?

Personally I had some reasonable results with it for some movies (lots of clear talk, less music)... alas, without any punctuation but precise timings.

Is such output acceptable for upload?

At least when there is nothing else available it could be a start for others to proof read.

[scooby007]

Everything that can be discussed and said about the subject has already been discussed here: viewtopic.php?f=1&t=1969

[0pener]

has already been discussed here: viewtopic.php?f=1&t=1969

Been there, read that Nah, does not really answer my question other than the general tendency there is to shun any machine-based things, so I hope you don't mind coming back

Just to be clear, there is:

a) machine-translation of written language (your thread's topic)
b) speech recognition (my topic)

These are to different things even though both are done by computers.

Don't get me wrong - I have no intention to flood OS with crappy transcriptions However, if there is no subtitle at all, I think then a STT-based sub shouldn't be categorically ruled out.

Wouldn't it be at least acceptable with the flag "Subtitles are machine translated"?

[SmallBrother]

a) machine-translation of written language (your thread's topic)
b) speech recognition (my topic)

These are to different things even though both are done by computers.

They are indeed two different things, but what is valid for A, is valid for B as well:
- Linguistically poor results
- Any technical aspect important for subtitles is not taken into account.

Don't get me wrong - I have no intention to flood OS with crappy transcriptions However, if there is no subtitle at all, I think then a STT-based sub shouldn't be categorically ruled out.

The argument "better something than nothing" is not valid:
- The majority of users are frustrated rather than happy.
- Any subtitle which is first, regardless the quality and regardless the machine translation flag, will 'eat away' most downloads. Making a decent subtitle will take (much) more time, be later, and thus serve only a handful of 'left-over' downloads. In reality, human subtitle creators often don't bother anymore. This will decrease or even zero the chance for a decent subtitle.

Wouldn't it be at least acceptable with the flag "Subtitles are machine translated"?

So, as far as I am concerned: no.

[0pener]

Dear SmallBrother,

First:
Quite frankly I disagree with you, at least partially... I still think something is better than nothing but that's your ship here, you explained it well and I stick with your rules

Didn't have time to respond earlier (also didn't want to look like back-talking). However, what is the flag "machine translated" good for anyway? I mean generally, what is that meant for?

Second:
It just came to my attention that someone has obviously that disabled flag for me anyway.



Just for me. Wait, let me rephrase that: BLOCKED JUST FOR ME!

Would someone have the courtesy to explain this, please? Who did that and why and why wasn't I informed about that at all? (Honestly, such clandestine moves make really angry and I think any decent person can understand that.)

[SmallBrother]

I still think something is better than nothing

I am sorry, but my argumentation is not based on personal opinion, but on facts.

However, what is the flag "machine translated" good for anyway? I mean generally, what is that meant for?

That flag exists for two reasons:
- Back in he days, machine translations were allowed site wide, but had to be marked as such.
- It still exists, so unmarked machine translations can be reported.
But I think I feel where you are heading to. That flag existing does NOT imply the 'right' to upload machine translations.

Just for me. Wait, let me rephrase that: BLOCKED JUST FOR ME!

As far as I know, this is not for you, but for certain languages, by request of admins of those language sections.

Anyway, your complaint gives me the feeling that you want to upload machine translations/speech-to-text subs anyway, despite you are saying "I stick to your rules". Am I right?

Would someone have the courtesy to explain this, please? Who did that and why and why wasn't I informed about that at all? (Honestly, such clandestine moves make really angry and I think any decent person can understand that.)

As I said, this is not something against you, but against machine translations.

Please note that OpenSubtitles is not a democracy, but a private house.

If I come to someone's private home and the person living there tells me to take off my shoes before entering, then that's simply the way it is and there's nothing "clandestine" about this. I would not argue that I don't agree, that my shoes are not dirty and everybody likes my shoes. And even if I would be the only one being requested to take off my shoes, that's the full right of that person. I would not argue that I should have been informed and I would not demand reasons and evaluate their validity. I would either leave (angry or not) or take off my shoes.

I think any decent person can understand that.

[0pener]

flag existing does NOT imply the 'right' to upload machine translations

Ah, okay then, seems I got it wrong there. Thanks for the clarification!

not for you, but for certain languages, by request of admins of those language sections

Ah, again. Sounds reasonable but I have not seen that anywhere else until now and I had another dutch recently... hence my surprise. (but then, perhaps it was disabled and I simply remember it wrong)

OpenSubtitles is not a democracy, but a private house

Don't get me wrong but this is incorrect: A more fitting analogy would pub or restaurant. Besides, as a newcomer I must say it can be hard to figure out your rules (OS has much room for improvement in this regard, how about a wiki or similar?). Anyway, your rules shall apply!

Many thanks for your answers/explanations and your time (to both of you, of course)! Very much appreciated! Happy weekend! <3

PS:

feeling that you want to upload machine translations/speech-to-text subs anyway

Nope, I want to but you forbid it, so there is not much room for interpretation - I won't. I just wonder if I'd have ever time to give my STT-subs a proper check other than just watching it. BTW, they are sometimes pretty, sometimes not (and the unpretty ones I delete myself).

[SmallBrother]

OpenSubtitles is not a democracy, but a private house

Don't get me wrong but this is incorrect: A more fitting analogy would pub or restaurant.

What I meant with my analogy is that some 'house rules' apply. But yes, you are right, a pub or restaurant would be a better comparison. After all, basically the door of OpenSubtitles is open to everyone.

Besides, as a newcomer I must say it can be hard to figure out your rules (OS has much room for improvement in this regard, how about a wiki or similar?).

Again, you are right. Except for a few spread forum topics about 'rules' (like that machine translation notice and a topic about credits), there is no real summary.

Advocate of the devil: not all language sections are the same in ' behavior' (user interaction, upload nature, subtitling guidelines, etc.). What may be a problem in one language section is not a problem in another one. And some things are not simply forbidden or allowed, but just better or worse. So any global summary of site wide rules would be either too tight or too loose. That's why I (as admin of the Dutch section) wrote some stuff about best practice, dos and don'ts in the Dutch part of the forum, but not in the English (general) part. Also, reality is that the majority of users do not really use the forum (or would search for a link about site rules). Many admins will just approach users directly and communicate about some problem if it occurs.

But yes, maybe we can think of some summary valid for all language sections, even if that would be incomplete or too loose in some cases.

Many thanks for your answers/explanations and your time (to both of you, of course)! Very much appreciated! Happy weekend! <3

And you, thank you for your kind understanding.

[pipporan]

But yes, maybe we can think of some summary valid for all language sections, even if that would be incomplete or too loose in some cases.

My perception is the same of yours: a few users navigate the forum.
BUT for that minority of users, i think it would be a very useful thing to have it explained why machine translations are not allowed, clearly and once for all...a post universally valid for all languages (it could be translated in other languages/section too).

A lot of users, even old and moderately experienced users, get confused by the "machine translation" check-box, and they think it's a legitimate option like the others.
And for sure it's one of the most time-consuming and less satisfying jobs for admins, to chase after every automated translation, check its quality and often disable the sub, and above all to explain to a lot of users why their subs weren't acceptable, every single time.

[SmallBrother]

i think it would be a very useful thing to have it explained why machine translations are not allowed, clearly and once for all...a post universally valid for all languages (it could be translated in other languages/section too).

True, at the moment things are a bit scattered, incomplete and too extensive at the same time and (therefore) unclear. Let's discuss this in the admin section.

And for sure it's one of the most time-consuming and less satisfying jobs for admins, to chase after every automated translation, check its quality and often disable the sub, and above all to explain to a lot of users why their subs weren't acceptable, every single time.

I totally agree.