This website uses cookies. Find out more or .

Questions and Answers

Question

1
point

I've a question concerning how progress is computed with your app.
As we all know, when we master a 'level' for a word, the word will sleep twice as long as the for the previous 'level'. But can a word ever move back one level, or at least stay at the same level after completion?

This is question is arose from my learning experience. I generally have no problem with the 2,4,8 days level. But some words are harder to remember, and when I come to the 16 days interval level, they suddenly look like if I saw them for the first time. Of course, the app will help me repeat them. I will have no problem to master this level, as to master a given level, you just need short time memory.
So the word will move to the next level, sleeping now for 32 days...  But actually I did not master that level,

I did not remember that word at all when I was asked for it after 16 days.
In my opinion this is a bit of a flaw in your model, you only need short term memory to master a given level, and to move from one level to the other, as for each given level (short or long time interval), the words will be asked within a short period of time in all four modes (which is quite easy).

I would like to find a way to better train the long term memory. So I am wondering if it would be possible to implement some kind of 'penalty' system, that make the word stay in a given level for a second/third time, or even go back one 'level' for the really bad cases. This penalty system would focus on how well you remember a word the first time you see it after a (long) period.

I don't know of there is an automatic method to implement this, I think it coul be quite ardous, not enough statistical values. But also a solution with a pop up would do.
Before a word goes to sleep for a period of time, you could just pop up a message saying:
"World should now go to sleep for xx days: yes  no downgrade".

yes             --> sleeping time increase by 2x
no               --> sleeping time not increased
downgrade --> sleeping time reduced to 1/2

In order not to be too invasive, this message should be displayed first from level 3 or 4 on. And of course, this behavour should  be a user option.

I trying to give you as much feedback as possible and make your application day for day a bt better ;)

Kind regards

  
This question has been viewed 22 times.
Answers
0
points
Thinking about it you are right, we should not move on with the sleeping time if there were obvious difficulties in passing the current training.

After some thought we are considering the following:
Before sending to sleep we will check the way the card was trained. If during the current training session there were some \'Do not know\' the idea would be not to increase the sleeping time further, rather keep it as it was and make sure that the training modes which included some \'Do not know\' would be trained again (those modes which passed correcthy without any \'Do not know\' would not be trained in that repetition.
Should a user have a problem again during the repetition of a word then we can even go further and reduce the sleeping time by half.

We think that a pop-up message as you suggest would be quite interfering and prefer a \"automatic\" method - think that above described method would address your issue and even would be training mode specific (thus avoiding to show things the user would anyhow know). What do you think?
Answered 157 months ago by klaus
0
points
Hi folks,

I like the suggestion & prefer the \"automatic\" slowing down if I keep having trouble after it wakes from sleep!

-Rebecca
Answered 157 months ago by Lush40
0
points

Hi Klaus,

Thank you for your answer.
Yes, I think that your proposal is the minimal functionality that could be implemted. You can implement a more sophisticated solution in the future if necessary. I also agree that for this \'trivial\' implementation there is no need for user acknowledge.

There is an other point I was thinking about, in order to increase the \'long term\' memory. To complete a given level the apps asks a world in all four modes. This happens in a quite short time, as wiil be asked those words each time I am training. Wouldn\'t be an idea to also randomly insert some short pause in this process (not too many, one or two). Those pause would be increased as the user reaches a higher level.
Here an example (X=repeat (mode is random))

Actual method: XXXX                    <-- T -->               XXX                    <-- T -->               XXXX
Proposal         : X      XX     X        <-- T-?-->              XX      XX          <-- T-?-->              X     X     XX

This of course works only for \'higher\' levels.
Also a slight difference would be that the card does not go to sleep for a fix time (T=16,32,...), but for this time less the sum of the \'pause\' (which should be about T/4).

I think a possible \'tivial\' implementation is:
-- only one short-pause    :   T/4
-- always the same cycle:   XX <- T/4 ->XX         <-- 3*T/4 -->             XX <- T/4 ->  XX .....

This should not be so complicated to implement, and I am conviced this would greatly help the users to assimilate a word.

Again I am asking for your opinion and other users opinion.

Answered 157 months ago by ribenguniang
0
points
May be the explanation given before was not clear enough.

- The proposed system will be differentiating between modes. So if you had slight difficulties when checking a word in a certain mode, then after waking up it will only show that more difficult modes. Additionally the sleeping time is adjusted (and not automatically increased).

- A card either will be sleeping or waked up. If it woke up, it will be necessary to be trained in all of the relevant modes. As you may know the training sequence is random - that is both the sequence in which cards are shown is completely random, and also the mode used is random (as long as they are different that the \"Initial training mode\".

- Something I did not mention here before: Based on input form other users (made about 6 weeks ago), we already have prepared to include a certain correction on the sleeping times which takes into account the overall difficulty the user had with a given word (that is the whole history the user has with training that word in the different modes and with special consideration of the the last round of training). The correction time  can vary between 50% and 150% of the \'original\' sleeping time.

Implementation:
The above method of calculating sleeping times and handling modes which were considered \"difficult\" will be implemented on our servers - it will be done in such a way that it will be working with the newest versions of our iOS, Android, Desktop and Java Phone applications - But! the information within the app on how long a card still is sleeping and also the list of cards memorized only will be accurate after an \"Update Cards\". Informations given at the time a card goes to sleep will be right in most cases but it can not be assured.

The advances method method with adjustments will be an option users will need to set in their settings page (online). We expect that we can finish the last test of that option within the next days and provide that option by first of July.
Answered 157 months ago by klaus
0
points

Hi Klaus,

Well, this is really a good news, because I think that the \'long term\' memory is not trained very well at the moment. I am really excited to get the new version, as actually I am \'intensively\' leaning, and I have quite a lot of words in the 32 days level, I have not really assimilated yet. I need this new \'brake\' feature (as Rebecca calls it).
I just wonder if the 150% is really realistic, in the \'32 days\' level it would meand a word would sleep in a given mode for 48 days, quite a long time actually.
But let us first enjoy this new feature, we can still do it better later with some experience.

Thank you to take so much time answering my questions :)

Answered 157 months ago by ribenguniang
klaus lautet:
As for the 150%, that would only apply if the card is trained at least in 3 modes and all of the tests passed without any one of them being considered "do not know" even once.
ribenguniang lautet:

Sorry this cannot be, it cannot be a longer waiting interval if you don't know anything. 50% would be in the case you don't know anything, while 150% would be in the case you know all modes, so actually the normal case, and this is much too long. I would say 33% to 100% is a better proportion.

Actually I find that the intervals in higher level are already now too long. I wonder if you could replace the linear serie 2,4,8,16,32,64 with a non-linear serie,  like 2,4,8,12, 24, 36.
This would give the user a quicker 'success' feeling.
In the actual solution it takes: 2+4+8+... = 126 days until a word is marked as 'memorized', while with the porposed non linear approach one would merely need 86 days, nearly 1/3 less.

Combined with my porposed solution for Repeat 'known words' (not yet commented by you), we could reach quite an optimal learning process.  

klaus lautet:
The 150% would only apply if you have Never! chosen a "not know",(never meaning over all the learning history of that word) so it will only apply for those cases where you actually already know the word.

The doubling of time for next sleeping period is a "standard". In any case those time periods will change quite strongly form case to case based on the planned algorithm.
0
points


Hi Klaus, this too was a concern of mine as more difficult words I have been learning I kept forgetting and really needed the sleep time to be reduced so I saw them more often. So I\'m happy to hear this is being changed.

That being said I am wondering if this has been implemented yet and if not how soon do you think the new version of the software will be ready?

Thanks Klaus!

Answered 156 months ago by quinntree
klaus lautet:
We are currently working on the Advanced algorithm. It still may take a couple of days to make it available (hopefully less than one week) .
Once available users who would like to use the advanced algorithm can switch that on using the Settings of this website. The handling of the sleeping time will be made on the server so there will be no need to change anything on the devices.
0
points
Hi Kaus, I am really eager to start using this new system and was wondering how much longer till it is implemented as here you said it should be no more than 1 week away, Now it\'s been over 3 almost 4 weeks I think?
Answered 156 months ago by quinntree
0
points
Hi we are nor life with the new sleeping algorithm including the new function which repeats specific problematic training modes with words.

As it is a mayor change in the way cards are handled we still are in \"Beta\" mode on this. we have been testing it as thoroughly as we could but having thousands of users with potentially different situations we just want to \'warn\' that it may still have problems.

To enable the new function, please go to the Settings page and select the Advanced Algorithm there.

The applications will handle the modes and sleeping time correctly - but as it is calculated on the server it requires frequent \"Update Cards\". And as stated also on another Q&A seems that the calculation of sleeping cards on the App is not correct.
Answered 156 months ago by klaus
0
points

Great!! Thanks I\'ll give it a test  :-)

Answered 156 months ago by quinntree
0
points
Yes! From now on I will train only with the Advanced Algorithm (I hope) !!!
Answered 156 months ago by ribenguniang