1. Forum
  2. >
  3. Topic: Duolingo
  4. >
  5. Let's do subtitles!

https://www.duolingo.com/profile/J_rob01

Let's do subtitles!

I think part of the long term development plan for Duolingo needs to include video subtitles. My experience is that listening to native speakers is vital to learning a language on more than an academic level. There would be both a dictation and translation component, probably not done at the same time. Can you imagine a little drop down menu where you could get subtitles in any language for any YouTube video?

December 5, 2012

39 Comments


https://www.duolingo.com/profile/saito200

This idea is AWESOME! Imagine being able to do it with a very short scene of a memorable movie, for instance. First you hear it, then you have to type in what they said. For me, one of the most difficult things on learning a language is going from knowing "the rules" and some vocabulary to actually understand native speakers. I am Spanish, using English in daily life for years, and still have some problems to understand people from the UK.


https://www.duolingo.com/profile/davidzweig2

You could take a look at this extension we made for studying languages with Netflix:

https://chrome.google.com/webstore/detail//hoombieeljmmljlkjmnheibnpciblicm?utm_source=Duolingo%20S


https://www.duolingo.com/profile/Pierre-O

It does exist on YuTube already. Sometimes you have a "CC" button to the left of the quality button (the weel).

You can take a look on this vidéo for instance : http://youtu.be/oN5tdMSXWV8 (A very smart speech by the way :) )

How it works : YouTube transforms the audio to text, this is for the native language of the vidéo.

Here is a blog post to explain the support of 6 european language : http://youtube-global.blogspot.pt/2012/11/youtube-automatic-captions-now-in-six.html

Then, this audio caption is automatically translated to the desired language. http://youtube-global.blogspot.pt/2008/11/auto-translate-now-available-for-videos.html

And finally the crowd can improve the translations.

I agree that it can be a huge value add for DuoLingo to have this feature. I read on quora that to learn a language, it is the best way to listen to a movie and repeat the sentences loud. http://www.quora.com/Learning-English/How-do-I-improve-my-English-speaking-skills-in-a-very-short-time

But As Luis already worked with Google (ReCaptcha), I wouldn't be surprise that it is already in the pipe :)


https://www.duolingo.com/profile/J_rob01

This Quora thread is interesting. I'm going to try the repetition technique. Thanks for posting!


https://www.duolingo.com/profile/merveaydin

Just check viki.com , It is a great project and people are willingly translating subtitles of many serials and movies, maybe Duolingo and Viki can work together. It would be so nice!


https://www.duolingo.com/profile/Beppe

Yes, videos spread a lot faster than books and their only limit is that spoken language is more difficult than written words ... worth a try


https://www.duolingo.com/profile/J_rob01

This is pretty neat. I wonder how accurate their translations are?


https://www.duolingo.com/profile/merveaydin

Each channel has an admin who is in charge of all other languages' admins. And each languages' admin is checking his/her subbers and segmenters. Segmenting and subbing interfaces are easily manageable by anyone. It's pretty easy to translate while watching the show, without stopping the video. The community is really paying attention (mostly consists of fans of the shows or actors) to translate properly and on time. I have never experienced any misunderstandings even about the specific cultural phrases or figurative meanings. They give explanations in parenthesis for idioms. Especially, popular shows have pretty accurate translation, because native speakers are translating their countries' shows to English, then other languages' teams rely on those subtitles. But of course all the system relies on its volunteer subbers. If someone sees a wrong translation, he/she can correct it. And if someone is really bad at translating, channel admin can ban him/her from translating. I guess admins can lock completely true translated lines. But there isn't any strong true/wrong voting system like we have here.


https://www.duolingo.com/profile/Fredecumber

This Viki.com is only for Korean and Taiwanese movies though...it would be amazing to see some site like that but more internationally oriented


https://www.duolingo.com/profile/merveaydin

Ah yes, before the time that I wrote that comment and before its new design, Viki was claiming that it will include more shows and movies from around the world. But now it is obvious that they are focused on Korean and Taiwanese shows. However, considering that all shows are translated in English, it can still be useful for users who are into such kind of entertainment.

And yes, it would be wonderful to have a more internationally oriented website with the same purpose. I guess it is all about copyright issues which was discussed in Viki's discussion forums. Viki aimed to include shows from diverse nations but only copyrights of Korean shows are easier to deal with.


https://www.duolingo.com/profile/Beppe

I am thinking that there are a lot of video on the net that can't circulate because of language barriers. Subtitles alone could help a lot and subtitles could be translated ... Difficult? Yes A lot of videos are amateurish and have bad sound quality but there is a lot of videos (speeches actually) produced by professors or universities for educational purposes that could be a lot more clear to transcribe. Duo could maybe set up his YouTube channel with material like this


https://www.duolingo.com/profile/Mephistopelus

We could even start by watching embedded videos from Youtube, and editing their automatically generated subtitles. If only Google would let Duolingo read and edit them.


https://www.duolingo.com/profile/xoviat

I believe I suggested this about a month ago, and Luis said that "we're working on this." However, I wouldn't be surprised if this feature was a long way off. That said, I still think that it is an excellent suggestion, given that the robotic speech from the lessons can only improve one's accent so far...


https://www.duolingo.com/profile/J_rob01

Great minds think alike!


https://www.duolingo.com/profile/mitcorb

I can attest to my low listening comprehension skill at this point in learning Spanish. I always have to use the slower button. Perhaps someone further up the learning scale could handle it.


https://www.duolingo.com/profile/Beppe

That could be a great idea: it is true that only advanced learners or native speakers can understand (and write down) the speech, but in my opinion some tool should be developed also for them.Nowadays Duo is mainly for beginners , but tomorrow?


https://www.duolingo.com/profile/ritagoel

I have been thinking on the same lines and trying to watch small videos with subtitles. I can understand the written word much better and I find I can understand only a little of what is spoken.Not to say that I understand the written language perfectly.


https://www.duolingo.com/profile/FreeXenon

Completely agree. Having more avenues for language immersion is good.


https://www.duolingo.com/profile/celebrim

How about collaboration with Ted Translations? TED transcripts could be submitted to Duolingo for translation by the users.


https://www.duolingo.com/profile/corazonHeart

I think that would be cool! Viki.com (http://www.viki.com) is pretty cool, too.


https://www.duolingo.com/profile/CaptainOwl

Hi everyone, this is a very good idea here ! Video would be nice, even if I'm sure they already have an eye on it in duolingo's team


https://www.duolingo.com/profile/sct16kevind

do you get points for this


https://www.duolingo.com/profile/christian

This wouldn't work at all. Language learners have very limited listening comprehension skills. Also, you can't look up new words because you don't know how to spell them.


https://www.duolingo.com/profile/tylrmurphy

I think that with a tool to slow down the playback speed of the audio, it might be possible. We would definitely need to experiment with it though.


https://www.duolingo.com/profile/Beppe

Worth a try, I my opinion


https://www.duolingo.com/profile/christian

What about unknown words, though? Many languages have a chaotic phoneme-grapheme relationship, which makes it virtually impossible to figure out unknown words.


https://www.duolingo.com/profile/Beppe

In that case we should obviously have a broken transcript that someone else could complete ....


https://www.duolingo.com/profile/J_rob01

I think by combining multiple inputs for different users you could come up with an accurate dictation in the same way Duo does translation now.


https://www.duolingo.com/profile/survivance

Native speakers would first transcribe 10-15 second clips. Multiple native speakers would transcribe the same clips before the system put confidence in the congruent sections of the transcript.

Then learners would do their best without the transcript, but could maybe refer to it by sacrificing a heart. idk

I see the tensions between delivering the best learning environment and outputting the most accurate and lucrative translations.

I have no idea how accurate or lucrative video translations are or will be over time, but I can imagine translating classic, canonical film clips to be personally rewarding and motivating.

Here's to Duolingo building the best introductory language system we humans have ever known :)


https://www.duolingo.com/profile/YvesLM

It would be challenging but there are many alphabets which are very close to being phonetic and would be a great place to start; Spanish, Indonesian or Swahili for example.


https://www.duolingo.com/profile/guidocalvano

Slower playback should be a completely solved problem. I remember the music making software I used to play was capable of these things. Creating an audio file that plays slower is trivial and countering the drop in pitch is also solvable. And there are probably already tools that do this for professional sub titlers.

Recognizing where sentences start and stop will be the hardest. And detecting words in the sentences will also a be tricky. But sub titleling can be something for the more advanced users right?


https://www.duolingo.com/profile/gerry11111

i respectfully disagree. i am an advanced learner of spanish and would rate myself as capable of translating most dialogue in spanish movies. i'd be more than willing to do so as long as duolingo includes the capability for native speakers to grade and correct me.


https://www.duolingo.com/profile/jess.fritze

I agree - I think this would be great!


https://www.duolingo.com/profile/christian

I doubt there are enough advanced learners to pull this off in meaningful numbers.


https://www.duolingo.com/profile/dvdtknsn

It's a fair point. It might be interesting for Duolingo to provide the ratios for beginners vs intermediates vs advanced learners.


https://www.duolingo.com/profile/sanda.golcea

Chris, I've actually seen this somewhere (I've searched for it, can't find it anymore, sorry). It was an app that would take a song from youtube, and would sync the lyrics on it - very much like you see it happen on spotify. Catch was that instead of you getting real time lyrics you'd get " _ _ _ _ _ _ _ " that you'd have to complete, and if you did not, it would stop the music and wait until you'd do it. Really sorry that I can't find it anymore. That was really good for even a beginner. However, I do question the legal aspects, and if the complexity would be worth it. I mean if you were spotify I'd definitely see it: they already have the music synced, they'd only have to modify that app to make you write and parse the strings, wait until you complete the line, then restart music.. but then again.. =)) haha sorry for writing soo much :)


https://www.duolingo.com/profile/gerry11111

this sounds like a cool app sanda; can anyone else help identify it?


https://www.duolingo.com/profile/christian

If the app uses existing lyrics, it's a completely different matter. The problem arises when no transcription is available to begin with.


https://www.duolingo.com/profile/CaptainOwl

I think that it is not a real problem if people make mistakes, everyone does. Duolingo have an algorithm that checks every answer and take the one that is given the most often by the duolinguist.

Learn a language in just 5 minutes a day. For free.