Feature Request: Most Decayed Skill Indicator
It would be extremely helpful if there were a way to quickly identify or go to the skill that is most in need of revision. I believe that would fit well with the teaching philosophy used by Duolingo and would most benefit those of use who have completed a tree and are working on improving or maintaining our proficiency.
The closest mechanism that is currently provided as part of the website is the 5-level indicator on each skill. However, there can be many skills that are at the same level, some of which have been very recently revised, and others that are on the verge of dropping to the next lower decay level.
As demonstration of the need for such a feature, I would point to the various script mechanisms that users have spent countless hours researching, developing, and maintaining (and repeating each time there is a change in the site), for example, the Skill Strength Viewer created by skelkingur. The problem with these is that recent changes in the site have meant that these mechanisms are not currently able to provide an accurate measure of skill strength (as shown by the 5-level indicators on the site).
An alternative to making a change to the UI would be to provide a higher resolution skill strength indicator via the API, which we could read via scripts. There would be no need to officially support or document this, just leak word to us. Please.
"DuoLingo tree trimmer" with the
duo.old_web_url_whitelist=["^.+"]whitelist workaround let's you filter 0/2/4/5 strength bars.
"DuoLingo skill strength viewer":
a) Shows skill percentage (concrete numbers).
b) already works for the new scala portal
c) but not all grammar skills are added for my EN-PT tree: https://www.duolingo.com/comment/24416411/URL-vocabulary-overview-backend-stream-is-missing-4-verb-skills-words-list-incomplete
Use both for your fun until new features are added the next years.
There is also a stylish replacement (Scala web portal compatible) to just hide GOLDEN skills instead of the "tree trimmer"; I would have to search for 2-3 threads, if you are interested.
Quote 1) The problem with these is that recent changes in the site have meant that these mechanisms are not currently able to provide an accurate measure of skill strength.......
Quote 2) I would point to the various script mechanisms that users have spent countless hours researching, developing, and maintaining (and repeating each time there is a change in the site)
Update: AHH, I should have definitely read your complete text the first time :-)
I think (these are my tests) that the old vocabulary API ("Words" tab) is bugged and "last practiced" user database column is NOT always updated regulary....or not all NEW words are being added.
Well, but why would the "strength viewer" not be accurate when the API even shows 0.0-1.0 strength per word???
It just uses a different - not 100% DuoLingo strength matching - calculation.
Then the API is different than their Skill web UI strength algorithm?!
And how can you even trust a "Duo strength calculation algorithm" if the user database for "learned words" is defective and strengthen exercises are not taken into account for several (massive!) words???
If the underlying user words DB has it's problems, then nothing selecting words and calculating words will 100% reliable work!
BTW: I love the "Dead words" 0 strength row. What the heck??
I have not been away for 3-6+ months! I practiced daily!
Like you, I would really love seeing a DuoLingo technical (backend core) team lead and senior developer take action and comment on the long-time known issues above:
- what they have (or have not yet) changed in the Scala web portal for those underlying backend data services
- why the API is still compatible with lekz's Android "Review flashcards" app!!!
- or if they are simply reusing old Python backend services at the moment because they run out of time (pardon me that I am not aware how deep a "re-engineering" goes at the first step)
- publish a good summary about actions planned in 2018 on their blog: http://making.duolingo.com