Stop flashcards from ballooning

daal

探花
I know this has been discussed before by people who understand it better than I do, but I feel it needs to be said again. I know that spaced-repetition does not work well if reviews are neglected, but what really does not work well is if the system wants you to review so many cards that you stop using flashcards altogether. The problem is particularly acute if you miss a day - often because you are busy, which not a great time to be loaded with extra flashcard work. The next day, you have double the number of flashcards due, and there is nothing to do but slog through it if you want to get back to normal amounts. From my perspective as a user, this is problematic as I don't have the time, and de-motivating. I don't want to be a slave to the dictates of my spaced repetition. If I'm busy for a day, I don't want my flashcard system to punish me for it. What I would like to see would be a simple way for the user to determine or limit the length of a flashcard session, and have the spaced-repetition adapt to my needs and not the other way around.
 

mikelove

皇帝
Staff member
We're working on that one - combination of triaging the cards you're most likely to forget + giving up on some cards altogether once you get really far behind. (which is the dicier part - basically at some point we have to conclude that the # of cards you want to learn + the time you're willing to devote to learning them simply don't match up)

In the meantime the simplest approach is to temporarily lower the 'points per day' in Card Selection to stretch out card intervals and then gradually raise it back to 100 as you catch up.
 

Shun

状元
One other idea might be the option of a more hand-holding spaced-repetition system similar to that of the quite popular "Memrise" app. It changes between studying modes automatically and quite intelligently, based on the learning success and time taken. Pleco could probably mimic their core system quite easily, just without the graphical bells and whistles and sounds included with Memrise, and even improve on it. Currently, for learning languages like Japanese or Russian, Memrise seems to be best app, but only with a pen and paper next to it to turn a Multiple Choice test, which isn't really sufficient, into a Fill-in-the-blanks test. A Pleco option which works in a way similar to Memrise but includes a lot more Fill-in-the-blanks or Self-graded would be quite something, I believe. :)

Pleco could change a score in the Pleco scoring system to the Learned threshold when a card is considered as learned using the hand-holding system, but otherwise keep the two systems completely separate scoring-wise.
 

HW60

状元
What I would like to see would be a simple way for the user to determine or limit the length of a flashcard session, and have the spaced-repetition adapt to my needs and not the other way around.
I fully agree! The main problem with spaced repetition is, that people finally quit. I had made a simple proposal some time ago (http://www.plecoforums.com/threads/getting-started-with-spaced-repetition.5211/#post-39933) where - within SRS - you can choose the number of cards you want to review. With a big number of flashcards you finally reach a point where you cannot review them all, so this is a natural development which needs support by pleco, and should not become a problem for the user
 

daal

探花
We're working on that one - combination of triaging the cards you're most likely to forget + giving up on some cards altogether once you get really far behind. (which is the dicier part - basically at some point we have to conclude that the # of cards you want to learn + the time you're willing to devote to learning them simply don't match up)

If you are going to triage the cards you're most likely to forget (maybe quarantine is better, as you might want to revive them later), you might also consider pulling cards that you are likely never to forget. Either way, I see that this might work to alleviate the pressure if the load gets too big, but I think it would be much better if it didn't get big in the first place. I would like to be able to say: "I want to do 100 cards a day" and not ever have to worry that the program feels I ought to do 200. If some are overdue, I'd rather the program manage that in the background, and just show me the 100 cards it deems are most important for me to see.

I'm sure you are right that there is probably a discrepancy between the amount someone wants to learn (everything) and the amount of time they are willing to put in (not so much), but it is not clear to a user how much they will learn in x minutes of daily study. All I know is that I want so spend a certain amount of minutes looking at flashcards. If I'm disappointed that I am not learning enough, that isn't a problem I'm going to complain to you about.
 

mikelove

皇帝
Staff member
Well that's the thing - a lot of people would complain if they're not learning enough; we kind of want to approach it from both the directions, the how much do I want to learn one and the how much time do we have one, and find a way to balance them algorithmically + at the same time make it clear to users what's going on. We've made a lot of progress (hired a retired physics professor to help us develop a statistical model to predict flashcard retention rates) but after this long we really don't want to rush it out. (and while we continue sorting that out we've made colossal amounts of progress in other parts of the system)
 

Shun

状元
Let me point out the following guiding principles:
  1. I don't want a machine (in this case, Pleco) to tell me when to study and how much to study. Human motivation usually depends on other factors than what a machine tells the human to do. And of course, there are unmovable time constraints, so Pleco should follow these.
  2. If I want to make the most of my studying time, I want an SRS to adapt to make full use of it, whether it's five minutes or 90 minutes.
  3. When I study, I usually don't want to have to think too much about the learning method. Most people just want "to study" and are very happy if the machine makes good choices for them on how to learn something, and how to build on it. This is something the machine can and should take out of the user's hands, if they choose to. That's what I meant with the hand-holding system.
I understand Pleco is more of a technical app, and that's good, but I do think it should offer something in that direction, because most users I've met with think "There are so many things I can do with Pleco, it's great to have all that power, and I'd like to take full advantage of it.", but then without clear direction, they lack the technical and practical knowledge to follow through. This experience is directly connected with user satisfaction, and this is what one should be trying to maximize.

To be clear, I'm very happy with Pleco, I'm speaking more on behalf of other users and based on the desire to contribute something. :)
 

Th41

举人
I am wondering, what does Anki do here exactly? Since it limits reviews to 100 and new cards to 25. I don't know if there is a more complicated system behind it, but is this regarded as ineffective?
I myself don't mind it so much, I have set my daily goal to 200 for myself and just quit the session when I have to do something else. But I see where the demotivation is coming from.
 

Shun

状元
I like Anki's system, too. I think it uses a formula to calculate repeat intervals based on the time between the current test and the last time a card was tested, and the number of times it had already been answered correctly. More of both makes the intervals longer. Cards to be learned are kept in stacks. You can set the number of new cards per day, but they don't add up when a day is missed. I think the previous stack is just replenished, but I'm not sure on that point.
 
Last edited:

mikelove

皇帝
Staff member
I don't think Anki does anything significantly different from what we do here; they just let you limit the number of cards to review and then show you whatever is farthest behind schedule first. Which we've repeatedly rejected because it doesn't provide sufficient incentive to catch up - if you stop the session after 100 cards then you're getting basically the same behavior you'd get in Anki with a 100 card limit, but in Pleco you actually have to manually stop the session and remain aware of how many more cards you have to review to catch up, rather than the session exiting normally and everything seeming OK.

Our current system was designed around the expectation that people wanted us to be more of a 'coach' than an 'idle time killer' - for people who still want us to serve the former role I think it works pretty well, but the changing nature of people's relationship with their mobile devices in the last few years means that an awful lot of our users would prefer something in the 'idle time killer' vein so now we need to find a way to accommodate both. (possibly explicitly through different modes / settings)
 

Shun

状元
Our current system was designed around the expectation that people wanted us to be more of a 'coach' than an 'idle time killer' - for people who still want us to serve the former role I think it works pretty well, but the changing nature of people's relationship with their mobile devices in the last few years means that an awful lot of our users would prefer something in the 'idle time killer' vein so now we need to find a way to accommodate both. (possibly explicitly through different modes / settings)

Great point! I totally agree. I think it would be rather easy (and perhaps fun :)) to implement an 'idle time killer' learning mode. Perhaps at the beginning of the session, Pleco could ask how many minutes one expects to spend in this learning session, then adjust the studying modes and number of vocab items accordingly.
 

daal

探花
Our current system was designed around the expectation that people wanted us to be more of a 'coach' than an 'idle time killer' - for people who still want us to serve the former role I think it works pretty well, but the changing nature of people's relationship with their mobile devices in the last few years means that an awful lot of our users would prefer something in the 'idle time killer' vein so now we need to find a way to accommodate both. (possibly explicitly through different modes / settings)

I'm not sure what you mean by coach (nag?), but my interest is certainly not in an "idle time killer." I view studying flashcards as an essential and effective method for learning a language, but certainly not the only method. I also need time to read lessons, do grammar exercise, talk with friends in Chinese etc. I also have a limited amount of time available to study. For that reason, I want to make sure that the amount of time I spend on flashcards takes an appropriate and sustainable amount of my study time. For situations where I am forced for one reason or another to miss a day, what would perhaps help me best, would be some kind of suspend option, in which the program would act as if the previous 24 hours had simply not existed.
 

daal

探花
This is just a bump, because I need to complain about this again. The number of due cards just keeps gradually increasing, and then I miss a day and bang, twice as many are due and it's more than I can manage. This is just not sustainable for me and I need a new way of setting up my cards so that I can avoid getting swamped. The pleco flashcard system is eating too much of my time and I need a sensible way of limiting cards.
 

mikelove

皇帝
Staff member
The best way to eliminate the impact of a single missed day would be to go into Organize Cards, search for all cards that have been reviewed at least once, then use a batch command to add 100 points to all of their scores.
 

HW60

状元
The number of due cards just keeps gradually increasing
The cards I think I will never forget again I EXCLUDE FROM ALL SESSIONS (in Card Info / Statistics). That reduces a little bit the increase of due cards. Still I think a SRS system with a fixed number of cards, as described in http://www.plecoforums.com/threads/getting-started-with-spaced-repetition.5211/#post-39933 would help against the problems with SRS. This would indeed be an idle time killer in the sense that whenever you are sitting waiting for something, e.g. for a train, you can review some of your due flashcards.
 

daal

探花
I don't really get why this should be described an idle time killer, as if you might as well be scratching your elbow. Doing some flashcards for a certain amount of time each day is surely beneficial to one's language skills. As it is now, I do mine on the bus, in line at the supermarket etc, because that's the only way I can get through them all. I think I'd learn better if I wasn't always rushed to finish an ever increasing amount of cards.
 

mikelove

皇帝
Staff member
Yeah, we're basically aiming to make the system respond to you; say how many you'd like to study right now and it'll figure out what to do with that time.
 

daal

探花
I don't really get why this should be described as an "idle time killer" as if one could just as well be scratching one's elbow. Studying flashcards for a certain amount of time each day is surely still a valuable element of one's language studies. As it is now, I do my flashcards on the bus, while waiting in line in the supermarket etc., because that's the only way I can get through the ever increasing number of cards. I think I'd learn better if I wasn't always in a rush to finish.
 

HW60

状元
I do not like if I miss a bus and have to wait 20 minutes for the next one. But I do not care if I have my a mobile flashcard system with me and can convert the idle time into useful time, e.g. reviewing flashcards. So maybe idle time converter is better.
 
Top