[WEB-783] Words learned discrepancy on (Web/Android)

Description There is a discrepancy between words learned on Android and Web.

Status: Low Priority / Unresolved

You can find more information on this bug, first reported in 2017, here: There is a discrepancy between words learned on the Android and the Web page

7 Likes

Yes, there is a discrepancy. I noticed it in my Danish 1 course (from English), because the Android app falsely claims that I have 193/196 words learned. I submitted a bug report about it in December. 2,5 months have past since then and the problem remains, no one bothered to contact me or fix it. How much longer do I need to wait? Half a year? Two years?

2 Likes

Hi @piotrmkucharski - We did create an internal ticket for this to be fixed in Oct 2018, but have not seen any progress on it yet. We are also aware this problem has appeared numerous times since 2017. I do not know when or if it will be resolved for good, but will let you know when I have more information.

1 Like

Thank you for the information, @MemriseMatty. I hope that it will be finally resolved in a foreseeable future.

1 Like

Pardon my interference, but from your description it looks like another well-known bug:

Finished courses show up as “unfinished”

Are you sure that this problem is a discrepancy in overall learned words count on different devices (total number in all courses)?

1 Like

Yes, you’re right. My mistake. But I suppose that these two types of bugs appear related. After all a course showing up as finished on one device and unfinished on the other is a special case of a discrepancy in learned words. So, if a Memrise team finally decides that the problem is worth their time and effort, fixing one should lead to fixing the other.

1 Like

Hi - We do have an internal ticket for this as well, raised in Nov 2018, but it’s current status is the same (No Progress / Unresolved). I will create another Known Limitation thread for this shortly.

3 Likes

Hi all,
I have a possible explanation for the bug.
Notice that the web version and the app version have
different difficulties. The App is much easier.
It seems that the App tries to compensate for that by repeating
words more often. So in order to have mastered a word on the
App you need to repeat it 6 times, it seems for the Web version
you must repeat the word less often.

To me it seems that each flashcard has a bunch of numbers attached to it.
How often repeated, how often failed, longest streak,
when next due, in long term memory etc.

Now, when computing “when next due”, it seems
to me that they use the number of repetitions.
And that works as long as you use only either App or Web.

Now if you mix it the following type of error appears:

You repeat it 7 times in the App.
Case 1: You repeat it the 8th time on the App:
It’s due next time in a week or so.
Case 2: You go to the Web version and repeat it the 8th time.
Memrise computes when it should be due next.
It thinks you repeated it 8 times on the Web due
and it is due in THREE Months.


-----
The same happens when you practice a flashcard on the web.
As it is more difficult to repeat it on the web it thinks you “know it”
after say 4 repetitions.
You go to the App and it thinks you know it only after 6 repetitions.
So words that were repeated 5 times
are counted in the Web version,
but not in the App version as finished.
Thus some courses are finished on the Web, but not on the App.
----
I think the top case is the most bothersome, as it marks words as known,
which are not really known to me.
-----
How to fix this?
First of all Memrise should be much more transparent on how the system works internally.
(By the GDPR, you have to tell us anyway how you are processing our data
and this falls clearly under this category.)
What I write up there is guess work. It is consistent with the reported bugs
and the information provided, it makes sense and is fairly simple.
So I would call it an educated guess.
But seriously, I have no clue what is really happening.
-------
Second of all Memrise, should keep track separately of the type of repetition.
Did the user spelled the word, just clicked on the right icon, or listen to it.
And a timestamp for each of those.
Think hard about all the information that you want to have.
Probably as many as possible!
------
Thirdly, there is a big issue with transitioning from the old to the new system.
But you can find a way. Just make it not too abrupt.
------
At last, compute “due to”, “in LTM” etc. consistently over the different platforms.
best
Till
PS: This editor seems to have a life on its own and changes the font all the time.

1 Like

I have a similar but slightly different issue. iOS and web show the same number of words learned. But totaling up the courses I have finished and the current words learned in my current courses, always gives 94 more words learned than what iOS and web show.