https://www.duolingo.com/mandrivnytsia

Different algorithms for different languages?

mandrivnytsia
  • 23
  • 22
  • 18
  • 561

I'm sorry if this issue has already been discussed before, but I haven't find it. I've been using Duolingo actively for two languages: Turkish and Italian. I've noticed some differences in the way the algorithm behaves for those two languages. First of all, my Italian skills decay much quicker than Turkish, but I guess that's because I've been testing out of many skills in Italian (due to having some experience learning Italian out of Duolingo) which I've never done in Turkish. What is more weird is the way the decayed skills are strengthened. I always use the general "Strengthen skills" button. If, let's say, I have 5 decayed skills, that's what will happen:

  1. In Turkish I will get sentences from my decayed skills in random order, at the same time I might also get sentences from other skills. It will take an average of 2-3 timed practice sessions to recover my decayed skills.

  2. In Italian In 90% cases I will need at least 5 timed practice sessions to recover those 5 skills. Sometimes it happens that 2 skills are strengthened at once, but it's rare. It also happens that even after getting 20/20 answers correct, my skill is still decayed (even if its strength before practice was 4/5). I've also noticed that I would always first get to practice the skills that are upper on the tree and only after strengthening them I get to practice the ones below them. So, to keep it short, the algorithm for strengthening the Italian skills seems to be much more strict than in Turkish and it feels a bit discouraging to me.

Have you experienced anything like that? Or do you think this difference comes only from my different learning style in those 2 languages (testing out of skills vs. doing all the lessons)?

1 year ago

7 Comments


https://www.duolingo.com/pentaan
pentaan
Mod
  • 25
  • 25
  • 23
  • 9
  • 848
1 year ago

https://www.duolingo.com/mandrivnytsia
mandrivnytsia
  • 23
  • 22
  • 18
  • 561

Thanks for the link! I keep both of my trees gold for almost all the time. What I've been wondering is why the regliding process differs depending on the language.

1 year ago

https://www.duolingo.com/pentaan
pentaan
Mod
  • 25
  • 25
  • 23
  • 9
  • 848
  1. Is the Italian tree much longer than the Turkish tree?
    My longest tree is much faster decaying than my other trees.

  2. Duolingo's staff is doing many so-called A/B tests. And it is really possible your account has been placed in the fast decaying group for Italian and/or in the slow decaying group for Turkish.

  3. Other decay factors:
    http://duolingo.wikia.com/wiki/Decay

  4. Try the method in my link. It really helps me to keep my longest and most difficult tree gold in only 1 to 5 strengthenings per day in the normal practice mode.
    It is very important to start with regilding the skills from the bottom of the tree to the top, instead of Duolingo's method (from top to bottom)

1 year ago

https://www.duolingo.com/Thomas_Wesley
Thomas_Wesley
  • 24
  • 15
  • 14
  • 12
  • 10
  • 8
  • 8
  • 7
  • 7
  • 7
  • 2
  • 7

It's an interesting question. So far I haven't notice a significant difference in any of the languages I'm learning (French, Japanese, and Hebrew from English, and English from French). French almost never needs re-gilding anymore (whereas the others often do), but I've been using that tree for about two years now.

1 year ago

https://www.duolingo.com/mandrivnytsia
mandrivnytsia
  • 23
  • 22
  • 18
  • 561

In my case Turkish needs re-gilding very rarely (while Italian decays very quickly), but I believe it's due to my learning style and to the fact that I've completed the Turkish tree a while ago while the Italian is only half complete. I just feel the way I need to refresh my skills in Italian is much more tedious than it used to be in Turkish before my tree became all gold.

1 year ago

https://www.duolingo.com/Gunnar388832
Gunnar388832
  • 24
  • 22
  • 20
  • 17
  • 15
  • 14
  • 12
  • 12
  • 12
  • 12
  • 11
  • 11
  • 9
  • 9
  • 9
  • 8
  • 8
  • 8
  • 8
  • 8
  • 8
  • 7
  • 7
  • 7
  • 7
  • 7
  • 7
  • 7
  • 6
  • 6
  • 6
  • 109

I would be very surprised if there is no language dependence. Not sure if it is due to A/B testing, language dependence in how the algorithms work (either directly programmed or due to some statistics from other people's answers), but Duolingo surely reacts differently depending on the language. I like playing around a bit and with Arabic => Swedish I got 7 % fluency from 81 xp!! That fluency percentage is even more insane than usual :-D, I have seen anything like that with other languages, where it would require in the order of 10 times that amount of xp to get the same percentage.

1 year ago

https://www.duolingo.com/Thomas.Heiss
Thomas.Heiss
  • 23
  • 15
  • 10
  • 704

To point 2)

The English-Portuguese tree behaves the same like your Italian tree.

You have to strengthen skill by skill manually, e.g with the help of the user script "DuoLingo skill strength viewer", but you should avoid the global "Strengthen skills" button.

If 5-10 skills won't strengthen after pressing the "Strengthen skills" button, it makes no sense to use it ;)
The sentences are NOT really mixed from several skills, but it locks vocabulary from 1(-2) skills only.
Usually I can strengthen 1, 2 or max 3 skills (but very very rarely!).

1 year ago
Learn a language in just 5 minutes a day. For free.

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