What is segmenting? Learning the technical skills of online translation at Viki (Part 2).

Jamie Henthorn's picture

Today is part two of my interview with a segmenter and volunteer trainer at Viki. Be sure to check out the first part of this interview, which was published Tuesday.


Today’s interview focuses on the training program this volunteer starts as well as the relationship that working volunteers have with Viki with regard to changing segmenting interfaces. More popular Viki dramas are shared through more mainstream streaming services, like DramaFever and Hulu. As such, volunteers might have their work seen by thousands. Viki, as a profit-making site, works with fans in different ways than many other online translation sites. As Viki updates its software it also challenges skilled segmenters and training academies to find the bridges that help volunteers adapt to new interfaces.


You started with Seg101, but now you head up another training program, the Ninja Academy. What are some of the choices that you made when you were putting together your learning environment?

I realized that Seg101 is one school dealing with a community of millions of people. It's not enough and they don't have sufficient resources to deal with so many people. It's a pity when there are so many people who want to help, but they have to wait a year. One of my staff was originally from Seg101. We got along together really well and had the same feeling that one school is not enough. We should create another environment to help educate more [volunteers] because the more schools we have the more resources we could gather and train more segmenters and teachers. And more teachers could train more contributors. It was time to think of another school and also teach what we think is crucial to segment well on live projects because I do a lot of that on live projects and give a lot of feedback. So I decided to bring my experience into my teaching.


I noticed you pull other software and resources to produce learning manuals and how to videos. I wonder how those choices were made. How did you choose your learning tools?

We just felt that we needed tools that were free and accessible to everyone because if we have something that requires downloading people might not know if it has a virus. If it's not free, like if we tried to print a book, they will be like "We are volunteers, why should we pay to work?" So, we wanted to find platforms that were universally used and universally accepted. My perception is that most people use Google Docs these days and then we could actually share it as published. Then anyone could just click on the link and go to those resources. So I think that's really quick and convenient and people can even download it as a Word document and just print it and and read it.


I heard a lot of feedback when I was doing the "Segmenting Diary." I want people to have access to that. I haven't written as much on the discussion board since it's more universally known that we have the Ninja Academy. I expect more people to try to reach out to our program to see a more organized, formalized, website description of everything that is more clean with videos and GIFs to show how to do do it right. It takes it up a notch. Picture's don't always mean a thousand words when there are videos to see how to actually do it.  


Overall, this MediaCommons survey is about technology and translation studies. How does the software that Viki uses help or hinder the ability to segment?

I've realized that back then the Viki tool didn't really show us how to segment. I don't think there was much information on how to use it. Just like the VikiU as usual, a one-minute video.


When you have the regional restrictions it's really hard. When Dramafever has the US/Canada license, but we can't see it we have to depend on one or two Koreans [on the team] who live in the UK. Then having them sub all twenty or thirty episodes is impossible. We are hoping that Viki in the future could further limit regional restrictions for QC. Because Drama Fever has it, maybe we could have some way to have access because we are a contributor, a power subtitler status when we have those people helping.


Have there been any issues with changes Viki has made to its software?

I'll discuss the changes I saw. It was from the timer that wasn't really user friendly. You had to time without being able to anticipate when there would be words again. So you had to really depend on your reaction time. It's more approximate and not as precise. With the new timer we could actually expect when a voice is coming and the waveform is pretty accurate. So when you hear the voice, you could rewind to the exact moment the sound starts in order to get the starting time for the segment. Using my XXL system, also it's much easier to just remove a huge chuck and just continue on.The old timer didn't allow us to use that. It's more on reflex and then no voice for a long time and when you hear a voice you have to chop and then a long time with no voice. You just kind of expect it like that.


[Another segmenter] gave a lot of feedback to Viki. She actually talked to the CEO of engineering of Viki at that time. When the beta segment timer with a half page waveform was used, figuring out which areas were bad and which keys and shortcuts we should keep from the old timer so that it would be an easier adjustment from the old to new. And then we came to the new timer, there was a point, it was only two or three days, everything was gray. They had a gray wave, a blue wave is more apparent, but with gray it is much harder to segment. So we made a huge fuss on the discussion saying which Viki errors they could improve on so they got a lot of feedback and they changed it back to the blue. They keep on implementing new ideas as we give feedback.


Viki also experimented with an innovative "auto-segment" feature. With this feature, the videos could be automatically segmented by their system. The purpose was to get channels that did not have many volunteers segmented immediately. However, this was an issue because segmenting requires intuition, which is something that their system could not replicate. This change received considerable amount of negative feedback so the official implementation was eventually held off.


Often Viki  will give you a survey. We write down what we actually think. There was a point when we couldn't actually put the video into parts. We had to wait a few hours or take turns. You would have to wait 10  minutes before you could enter the timer after a segmenter finished their turn. So we would waste a lot of time segmenting a video, maybe 3 hours or so, compared to when it's pre-split  into the six parts and six segmenters could segment at the same time. We could get the job done in an hour. Viki gives the Channel Manager the opportunity to self create parts now, which is really nice.


Is there anything the world needs to know about segmenting Viki and the tech used?

I think Viki incorporated a lot of elements that are innovative compared to other programs out there, like Aegisub. That one also uses a wave form, but it's not as user-friendly as the Viki one. The Viki one allows us to choose by parts so we won't be exhausted as with Aegisub, there isn't a lot of break time and the video is much smaller in the corner and the wave is much bigger. It's not as user friendly. [Viki] is simple, direct, and easy to use, which makes it much more interesting.


Megan McKittrick's picture

Great interview!

What I really appreciate about this interview is the fact that you've asked questions that get after so many different angles of this topic, making it an interesting artifact for researchers in multiple disciplines and sub-disciplines. As a researcher in technical communication, I find it fascinating to learn more about the work patterns of this unique group and how they employ technical communication documents, like manuals, to maintain their community. It sheds like on writing practices in really interesting ways. 

Jamie Henthorn's picture


I'm quite grateful to have had such a good interview subject. I'm interested in the technical communication here to, the number of drafts that many of these learning modules go through, and the usability testing that is done on documents. Likewise, and you may know more about this, the ways that part of the technical communication is the ongoing temporally-based feedback. 

JessJackson's picture

Again, Jamie, thank you for

Again, Jamie, thank you for the interview on this subject. It was something I wasn't quite aware of even though I did question from time to time, who were the ones that did the fansubbing and how did they go about it. Looking at it from a research perspective brings in a lot of different ways that fansubbing and segmenting can be utilized as a skill or a hobby as well as how it brings communities together. What is even more interesting is the aspect of volunteering as many are giving up their time to do this for others. Thank you for sharing.

Dr. Tessa Dwyer's picture


Thanks once again. The issue of licensing is really interesting, highlighting when geoblocking goes wrong – hindering the work of legal, volunteer translators. Also fascinating to hear how this segmenter appreciates the user-friendly nature of Viki's translating tools. Its something that seems really key to Viki's success and reach. Its such an open community.