Viki's standards being low af

[helenawang5717_57]

For me, I don’t look at VIKI and its community in a money-related way that this is a ‘paid-must-have’ community just because it is something some people do as their job (at other places) although I often got asked by people who are not part of this community why I even spend so much time in doing something I don’t get money for (that is usually the first question people ask when I mention I’m doing some volunteering for Asian dramas: How much money do you get for it? And when I said nothing they say: Why are you even doing it then? It’s no use to do it when you aren’t paid for it.).

I think of viki as something I don’t expect money from, and I think many other people do too. Volunteering is something people should do if they love doing it, and I think this is one great aspect of viki. I don’t think others should stop you from helping even though you are not paid.

I sometimes saw angry comments of viewers complaining about the ‘slow’ speed of new subs or that ‘their’ language was not available for them so I’d like VIKI to make it clearer for viewers that this page’s subs are mainly done by volunteers who spend their time for free to offer as many languages as possible.

I agree with you so much on this issue. I don’t know how many times I have seen complaints all over the comment sections and reviews about “why are there no subs, I paid for them”, “why did subs stop”, or “why does it only say thirty percent?”. I think when people first join viki, they should be sent a message or pop up screen about how subs work and that vikipass does not cover subs. Or viki should make it clearer on show pages where subs come from. I think this way, the comment sections will actually be filled with things ABOUT the actual show, and it will reduce the number of bad reviews just because of subs.
Another solution is that shows be released to the public when they are about 90% subbed. That way, there will not be so many complaints.

[irmar]

it happened several times that 2-3 weeks after the English main team already released several episodes for all other languages they said they gonna change stuff (like terms or sometimes whole sentences).

Unfortunately this happens because of the T.E.s not coming on time. There are very few of them, too busy , and we often have to do as best we can without them until they are available to come.
The episode is at 100% subbed and the first English editing done (by this I mean something that I, as an English editor, do immediately after the subbers: format, punctuation, spaces, italics, spelling of names, job titles, consistency of flashbacks), but there’s no sight of the TE.
The days pass, the other language teams are waiting to start, the other language viewers get impatient (and for once they are perfectly right!).
So what happens? We skip the TE. What else can we do?
The English editors edit without previous TE, asking the subbers or other friends wherever there is some strange thing. Sometimes we look at the script, if it’s available, sometimes we check translations from other sites to make sure we got the meaning correct because it may happen that what our subbers translated makes zero sense and we can’t even understand what it means.
Then we go for another round by the Chief Editor and we release. Telling the other language mods that whenever the TE comes, we will tell them of any significant changes: by which I mean whatever changes the meaning, not just a little word here and there.
When I’m in the team and this happens, to make the OL mods’ life easier, I send a list of every change with the precise timing of each, so that they can do it quickly and effortlessly.
If the team is lucky to have a CE knowledgeable in Korean (a rare thing, but there are a few), the quality is high anyway. But when the TE comes, she always finds things to change! Some may be unimportant (and I beg her to leave out those, since every single correction will have to be checked!), but some ARE important, they actually change the meaning or add an important element, so it would have been very bad to leave them as they were.
Of course, the native speaker TEs often mess up the English, so you always have to pass once more after them. You can’t leave them to be the last person who made changes.
Do you realize that in such a situation, the English editor has to do three passes?
Yes, it has happened to me that the TE came when we were already at episode 12 or so out of 16. And she went on a roll, working like mad and editing one episode per day with the goal to complete them all before the show ended. At the end of the day I was nauseous with exhaustion, having worked for more than eight hours on re-revising, updating the changed subs on our Google sheet and compiling a list for the OL with the minutes, the old and the new. It felt like forced labor. That’s how I spent my Christmas holidays this year. Ugh! I hope I never have to do this again.
I was surely happy because some of the changes were really important and necessary, resolving lingering questions that had remained unanswered. But how much better would it be if TEs came exactly when they’re needed, which is right after the subbers?!

[worthyromance]

If I am not mistaken, the pre-subbed dramas are the ones Viki shares with Kocowa, which requests pre-subbing as part of the deal.

I looked for the shows I edited which came pre-subbed, and I couldn’t find them on the 39 pages of the Kocowa section of Viki.

Another solution is that shows be released to the public when they are about 90% subbed. That way, there will not be so many complaints.

There are people who speak Chinese, Korean and Japanese who watch Viki. They don’t rely on subtitles for their enjoyment of the shows.

Do you realize that in such a situation, the English editor has to do three passes?

That’s so true!

[mirjam_465]

I never ever expected to be paid by Viki. I knew from the beginning it was an unpaid job and I’m okay with that. I like the fact that viewers can watch Viki without paying or without even having an account, unlike Netflix and the like where you get to watch zero shows without paying (with the exception of free trial periods, but that’s by far not the same).

I do think that, within reason, volunteers should have a say in how some things are being handled. After all, they are not only serving the so important viewers, they also ARE viewers themselves.

Some viewers seem to be extremely impatient. Maybe that’s partly because they don’t get the right information. Maybe their subscription should come with the warning that only English subs are guaranteed and that OL subs are just a bonus. And either way that subbing takes time.

[mariliam]

Hi there everyone,

I just wanted to clarify some points and concerns stated above.

As you all know some dramas are released with English subtitles, and of course, with segments included. These are, for the most part, from the content owner’s own available SRT files. These files always come timed (segmented) as well as subtitled and we’re able to upload them into our service under our Viki account. This is done intentionally so that you guys can see the subtitles were either uploaded, paid for or if they “came with the show”.

As we have a small content operations team uploading these SRT files to these episodes, we’re not able to make changes to these SRT files internally since it will take a long time and the team has multiple shows to handle at the same time. So what ends up happening is that they get uploaded as received.

We understand that some/most of them don’t meet our community’s current standards of quality, but keep in mind that these are given to us by our content distribution partners or made by professional firms (in the example of Kocowa shows) and that we upload as-is, as we have little time before moving on to another project. What I’m trying to say here is that staff is not creating segments manually, but simply uploading SRT files to episodes.

Now, the other thing I also want to make clear is that we’re not segmenting these shows manually nor using any other type of automated technology to segment these shows either. Our only AI technology available on segment timer pertains only to the waveform to identify speech and remove background music or noise, which is meant to help give better visual queues for segmentation.

We did do some experimentation back in the day with automatic segmentation but it proved to be a challenging project and it was dropped. This is no longer available to us as a service or option either internally nor externally.

I’m hoping this helps clear up some of the points raised above. More than happy to continue discussing this with you guys here or on a call if necessary. This is important and we need to find a way to avoid disrupting your experience segmenting, translating, and editing.

Thanks,
Mariliam

[cgwm808]

Thanks Mariliam. I knew there was no segmenting bot but I think the suspicion was raised because some people don’t realize how software can put together a soft sub file like srt format and a video – they’ve never gone through the exercise of synching the two!

[piranna]

This is important and we need to find a way to avoid disrupting your experience segmenting, translating, and editing.

Segmenters who are fixing or who have fixed presegmented content might be interested in contacting you. How can they contact you?

[worthyromance]

We understand that some/most of them don’t meet our community’s current standards of quality, but keep in mind that these are given to us by our content distribution partners or made by professional firms (in the example of Kocowa shows) and that we upload as-is, as we have little time before moving on to another project.

I second the thank-you reply for this information! This has helped clear the issue in my mind. A pre-subbed show or episode takes a day or two for our skilled volunteer segmenters, then several hours for a general editor to bring those subs up to Viki quality. Viewers then don’t need to wait a week, and in rare cases up to several months, to get the episode’s translation.

Knowing they arrived as-is from the content provider or professional firms gives me peace of mind regarding their translations. As a viewer, I noted that timing may be off and grammatical/spelling/punctuation errors may abound, but I have been able to watch and understand the show immediately. That’s a distinct advantage for viewers, in my opinion.

As an English editor, I now understand why the TE has about the same number of subtitles to edit as in a show that comes without subtitles. If our volunteer origin translators or their professional origin translators do the work, I don’t see a problem with that.

In order to answer complaints about timing and the improper English, before it gets fixed, are we allowed to state in the show’s comments section that this came pre-subbed by professional subtitlers, but we will fix issues as soon as possible? So many viewers complain that Viki should pay subbers for quick subtitles. If they understand that this is the quality of paid subbers, they may not complain in future.

[bozoli]

Thank you for the info. I have to say, I am thoroughly enjoying you engaging in our discussions.

Back to the point in question.

If we take a look at the original post, we see that @honeybuns came across segments which were not just a bit bad, but actually terrible.

Some segments start .3+s before the first hint of voice, some start after the dialog started. Extensions range from nonexistent to just short of 1s.

These are some very concerning gaps in quality, to the point where your average viewer (otherwise not interested much in quality of the subs, but rather invested in following the plot) starts to notice something is very much off.

Furthermore, @worthyromance writes…

A pre-subbed show or episode takes a day or two for our skilled volunteer segmenters

If I’m not mistaken, an average one hour episode takes 2-4 hours to get segmented de novo.

Compared to the 24-48 hours, this means segmenting is significantly extended (12-24x) for pre-subbed shows.

Subtitling, on the other hand, seems to be much quicker, as it is reduced to pure editing.

Here is my suggestion. Same as volunteers nowadays get Korean scripts to translate from or get them as help, why don’t you:

1. allow Viki’s segmenters to segment the pre-subbed shows by Viki’s superior standards
2. give SRT file to either volunteer subbers or directly to the translation editors, to input the translation from SRT into Viki’s own segments?

[sonmachinima]

If they understand that this is the quality of paid subbers, they may not complain in future.

Good idea, I hope they read the info then

[mariliam]

Hey everyone,

My answer to your individual questions below:

Segmenters who are fixing or who have fixed presegmented content might be interested in contacting you. How can they contact you?

I’m always available via PM and likewise, it might be beneficial to hold a small (or big) forum to talk about this on a call. How does that sound?

In order to answer complaints about timing and the improper English, before it gets fixed, are we allowed to state in the show’s comments section that this came pre-subbed by professional subtitlers, but we will fix issues as soon as possible? So many viewers complain that Viki should pay subbers for quick subtitles. If they understand that this is the quality of paid subbers, they may not complain in future.

This is a very interesting idea! I think that for something like this to work, we will need to set up a more transparent workflow when posting channels up on Project Finder or when selecting CMs. There are also a few of these shows where we’re not allowed to make changes to these translations, which then makes it easier in terms of communication to users. As you point out, it is not evident to users that these subtitles were not made by the community and that should not be the case, but that will require a lot of changes on our platform that are not scheduled yet! I’ll be thinking about this a bit more and will bring it up internally for discussion.

Here is my suggestion. Same as volunteers nowadays get Korean scripts to translate from or get them as help, why don’t you:

1. allow Viki’s segmenters to segment the pre-subbed shows by Viki’s superior standards
2. give SRT file to either volunteer subbers or directly to the translation editors, to input the translation from SRT into Viki’s own segments?

We’ve actually had this option before for various shows when importing subtitles, but in the interest of time, decisions have been made to upload with subtitles included as to not overwhelm the community. This is something I would love to discuss with you guys on a call as I can see a gap in how we are trying to “protect the community” and also creating more work or trouble with our “good intentions”!!

I love these discussions as they challenge a lot of the preconceived notions we have acquired over the years!

Cheers,
Mariliam

[piranna]

Thank you for making yourself available!

I think it could be a good opportunity for Viki and segmenters to discuss about it.

So maybe, on volunteers’ side, we could first inform segmenters about this topic and a future discussion with Viki is possible,
since it impacts their volunteering experience directly and also segmenting academies because people begin to doubt (like a volunteer in a post above) whether they should still enroll to this program or no, knowing the current opportunities for a graduate, that as a matter of a fact, have changed compared to 2-3 years ago.

The best way to contact segmenters, I think it would be to contact seg101 and nssa because segmenters all come from the segmenting academies and are registered there and also, to see with these academies.

I know that recently nssa has made a feedback survey for their graduates. Maybe some of their replies could be an indicator to understand the current hardships they encounter.

So to sum up, I think segmenting academies and their segmenters should be informed because it is about their present and future on Viki. This discussion should also include them.

[irmar]

an average one hour episode takes 2-4 hours to get segmented de novo .

Compared to the 24-48 hours, this means segmenting is significantly extended (12-24x) for pre-subbed shows.

That’s a gross exaggeration. I’ve done both and I think it takes more or less the same time, to correct than to make segments ex novo. Probably slightly more, if you are an experienced subber, but surely never ten or twelve or twenty-four times more!

If the pre-subbed show requires a lot of editing, for example twice or thrice as much as editing an average translation from Viki volunteers, then we do have to ask ourselves if this up-and-coming process is efficient and productive enough.

That’s also an exaggeration. Sometimes Ko-En (or C-En) Viki subbers are awesome, and they need very little correction. But some of them write in a really sub-standard English. Which is never the case in professional subs. At least those always make sense.
Professional subs have two main problems:

1. condensing (because of brevity rules)
2. localization, which loses all the local flavour
   and a third, which is not a “problem”, just a question of different conventions:
3. the different formatting.

The localization and the formatting, as well as any naming conventions, spelling etc. I usually take immediately care of myself, before the TE. So that the TE can take care of the missing things because of condensing, and can work quickly and efficiently.
However, all the simple little dialogues like this:

Where are you? - I’m coming right now" - Okay, don’t forget to bring beer. - Okay, see you in a while.

Those usually don’t need to be changed, and that’s a lot of work already done for us, it surely saves time.

[bozoli]

We’ve actually had this option before for various shows when importing subtitles, but in the interest of time, decisions have been made to upload with subtitles included as to not overwhelm the community.

Ah, so it wasn’t fast enough. Why am I not suprised?

I can see a gap in how we are trying to “protect the community” and also creating more work or trouble with our “good intentions”!!

Yes! But more importantly, it is a discussion of how to preserve high quality of Viki’s subtitles (segments and translations), while at the same time maintaining good speed (and sub-shiners at bay) and not overburdening the community. If we take a closer look at particularly @piranna’s posts here, they focus primarily on efficiency of the whole subtitle-making process, while trying to maintain the standard quality of the subtitles.

If one type of work pipeline doesn’t fit every project (whether it be direct upload of the SRT file or giving the translation to the English team) then perhaps it might be good to introduce a bit more flexibility. For projects where pre-segmenting and translations are good, of course uploading the SRT file directly is great. On the other hand, in @honeybuns’s case, we see that it would have been more efficient if someone noticed large gaps in quality of segmenting before uploading the SRT file. Because that team is going to have to now spend a lot of time fixing those segments.

I have no idea how many series per day and which percentage of series come pre-subbed to Viki, nor if the quality of segmenting is consistent throughout all the episodes of one series. Assuming the latter is true and the quality is constant, would it be feasible that a Viki staff member who is uploading the first episode (or two), checks segmenting in the SRT file?

bozoli:
an average one hour episode takes 2-4 hours to get segmented de novo .

Compared to the 24-48 hours, this means segmenting is significantly extended (12-24x) for pre-subbed shows.

That’s a gross exaggeration. I’ve done both and I think it takes more or less the same time, to correct than to make segments ex novo. Probably slightly more, if you are an experienced subber, but surely never ten or twelve or twenty-four times more!

This is a direct calculation from what has been told here by others. Of course, you might have different experiences. But it seems some pre-processed projects are harder than others.

bozoli:
If the pre-subbed show requires a lot of editing, for example twice or thrice as much as editing an average translation from Viki volunteers, then we do have to ask ourselves if this up-and-coming process is efficient and productive enough.

That’s also an exaggeration.

This was just a presumption, not a statement. Later it was stated that editing English of pre-subbed shows on average takes less than editing the translation of Viki’s volunteers.

A pre-subbed show or episode takes a day or two for our skilled volunteer segmenters, then several hours for a general editor to bring those subs up to Viki quality.

Which I have also acknowledged here:

Subtitling, on the other hand, seems to be much quicker, as it is reduced to pure editing.

However, all the simple little dialogues like this:

Where are you? - I’m coming right now" - Okay, don’t forget to bring beer. - Okay, see you in a while.

Those usually don’t need to be changed, and that’s a lot of work already done for us, it surely saves time.

Sure, I understand that. All in all, I’m glad not everybody has problems with subs coming from outside of Viki.

[sonmachinima]

I’d like to have some kind of additional info box in which subbers can write background/detailed infos and this box should be activated/deactivated per viewer in the way those who aren’t interested can just turn it off. Sometimes the subtitles get really long (way too long) because of additional translator notes and I think such a box would be a good way to keep subtitles clear and easy to read but also give background information.

[cgwm808]

@sonmachinima
Way back in 2011 when I was corresponding with the engineers when they first planned to add the sound wave, they were actively considering another line for the subbers to use on top of the screen to put translation notes and song lyrics but alas it was never implemented. I also proposed that they have the sound wave in different colors to show the addition of a different voice (which would have aided in the development of a bot segmenter!!!).

[anna79_9]

Queen Dugu Example (50 episodes drama) → english subtitle provided by the content provider (Btw, is it consider professional or not?).

13 segmenters + 2 Chief segmenters:

Segmentation correction for 1 part: approximately 1 hour or more.
Plus, Final verification by 1 Chief segmenter for 1 episode: approximately 1 hour or more (for me, 1.5 to 2 hours)

2 English/Chinese TE (Translation Editor): → MEANING!
First with 11 622 translation.
Second with 7 294 translation

→ Ok, here we are: This drama has 30 588 segments with approximately 32 segments by episode for credits and beginning/ending OST, so for 50 episodes: 1600.
30 588 - 1600 = 28 988 (In this, there are also all the OSTs that are not beginning and ending OSTs included, which I am too lazy to look for. Not translated by provider)
So TEs changes: 11622+7294 = 18916

18916*100/28988= 65.25% changes by TEs. So 34.75% translations provided are kept.

1 GE (General Editor) with 5423 translation
1 CE (Chief Editor) with 6511 translation.

It’s difficult to assess the time spent by TEs, GE and CE, but I think we can easily deduce that TEs has almost retranslated this drama.

Martial Universe example (60 episodes drama) → Provided by viki contributors.

9 segmenters + 2 Chief segmenters.

Segmentation for one part: approximately 30/45 min (little more for beginner)
If Final verification: approximately 30/45 min (for me, 45 min)

1 TE with 5670 translation.
2 GE with 3772 and 3237 translation
1 CE with 2008 translation.

→ So, here we are: This drama has 26658 segments with approximately 23 Ending OST and credits by episodes, so for 60 episodes: 1380
26658 - 1380 = 25278 (In this, there are also all the OSTs that are not beginning and ending OSTs included, which I am too lazy to look for)
So TE: 5670

5670*100/25278= 22,43% changes by TE. So 77.57% provided by viki translators are kept.

I can give you lots of other examples and statistics, which will prove that we spend more time correcting whether in segmentation or editing / retranslating, than doing it ourselves.

We can see that the segmentation and translation/meaning are affected a lot here, in time and changes. We can even wonder if it makes sense to keep this content… Unable to quantify editors time, unfortunately. (CE and GE are not significant, moreover, they are linked. More samples are needed for this to be significant.)

For segmentation, you guys should also know that segmenting and correcting segments are two different things. Segmenters are trained to segment, not to correct. This is why it can take more time to correct, or also more time for the Chief segmenter, who checks as Final Checker behind the segment correctors.

My personal feeling: Correcting is an horrible task, it’s difficult and tiring. It takes time and a lot of energy. Reason why I don’t do it anymore. I know a lot of segmenters don’t like to do it. I’m sure many TEs don’t like it either.

It’s obvious that we must find a solution for the collective well-being, and think of the whole community of segmenters, English editors. And for standards quality and accuracy provided for viki viewers.

I mainly speak of Chinese dramas. As for Korean dramas, I don’t think I have seen any of these. But they seem to provide translations by professional firms for Korean ones involved, which makes a lot more sense, and which is not the case with the CHINESE dramas. Well, if this is provided by professionals, they should be fired right now.

@mariliam

[piranna]

… forget about me, I already talked to Viki last call. If it was only me, we wouldn’t have this discussion.

I am a volunteer first, not a company, but I understand companies, not philantropist and having consumers.

Speed vs quality (& vs contributing: because some of us gave up on these presegmented and presubbed, that is segmenters or TE).

I also understand some editors and viewers for whom this system could be better because it is faster and for whom the focus is not the same as a segmenter’s. An editor focuses on subtitles, not segments (or it’s relative). And it depends if they are GE or TE.

I also understand segmenters and subbers and some editors (especially TE) for whom this system could be heavy.

If I have to tell my focus:

• Do we contribute without pleasure?
  Most of us will give up, which is happening.
  We will focus elsewhere.

• Quality and meaning: Viki can ask and talk with some editors or TE about the quality.
  Their counter on some presubbed dramas could be an indicator.

• segmenters. Between a show where they gain contributions, make their 500 segs, and where they take pleasure by segmenting themselves vs a show where they have to review segments that don’t respect the standards they have been taught, where they might take longer and gain almost nothing in contributions…
  And that could explain why segmenters don’t want to go there anymore.
  We also take pleasure in segmenting ourselves.
  Do we take the same pleasure in presegmented dramas?
  From 4 segmenters’ answers above, no.

• we’re a team and we take pleasure in contributing.
  That is the case for TE and Asian subbers.
  If it’s already done, where is the pleasure? Where can you meet TE and Asian subbers? I spoke with some and they’re pretty nice. They also love what they do.
  If we get presubs in most popular languages, only editors needed and only those who want to edit. What about subbers?
  For my language, not many people can edit and want to edit until the end.

What’s the difference between a bot and presubbed?
Where is Viki then? The original spirit?

[piranna]

Edit: the graphs and tables are up there.

If I can compare between these 2 dramas that have about the same number of ep.

Presegmented and presubbed vs. Done by contributors.

For a GE:

• -1.5k less in a pre (-20% edition)

For a CE:

• +4,5k more in a pre (+225% edition)

If we mix GE-CE:

• +3k more in a pre (+32% edition)

For a TE:

• +13k more in a pre (+230% edition)

For a segmenter:

• +15-60 min / part (x1,3 - x3 in time / ÷5 - ÷10 in contributions for a part of 100 segments)

For a CS:

• +15-30 min or +1h-1h30/ ep (x1,3 - x4 in time / ÷5 - ÷10 in contributions for a part of 100 segments)

People who edit more and so take more time:

• TE (×3 edition or ~3 dramas alike done by volunteers = ~1 pre)
• CE (x3 edition or ~3 dramas alike done by volunteers = ~1 pre)
• segmenters (x1,3-×3 in time or ~2-3 dramas alike done by volunteers = ~1 pre)
• CS (x1,3-×4 in time or ~3-4 dramas alike done by volunteers = ~1 pre)

On the contrary, if you’re a GE, you have less work (×0,75).

If we mix GE-CE (×1,32 edition)

(It is calculated for these 2 dramas only.)

[irmar]

65.25% changes by TEs. So 34.75% translations provided are kept. (…) I think we can easily deduce that TEs has almost retranslated this drama.

No, we cannot easily deduce anything. And I’ll explain what I mean.
There are many times when having a basis on which to work, even if I have to change a detail or two, is quicker than doing it from scratch. Because it gives me a first idea, and maybe a word that I would have to search for etc.
Even when I am Italian editor and edit my Viki Italian subbers, sometimes I stand in awe of a particularly good translation, and I think to myself “I would have never thought of that!”

I’m not saying it’s always like that. Sometimes the translation is plain wrong, and it’s obvious that a TE or GE has to change the whole thing. But let’s not forget the many, many cases when it’s not so. Contrary to Viki initial subs, pre-made subs are very rarely wrong in meaning. They may be incomplete, they may change the expression a bit too much, but 97% of the time the meaning is correct.

And, of course, sometimes the correction is only formatting or punctuation. If you give me an episode where I only change italics, breaks, and the spelling of character names, I will show you that I will have made at least 20% of changes. If there are many flashbacks and past scenes, even more.
I’m not talking theoretically, I’ve had to do this a number of times, for instance in The Package, Revenge Note 1 and 2 etc.

So numbers don’t always say the whole story, and they don’t mean exactly what we think they mean.