Mapping my Understanding of SRS to Pleco Flashcard Settings

[HW60]

Let me once more ask for a solution for the backlog problem. I think that it is important to accept, that this problem exists, as it is a natural result of SRS. With only a moderate amount of flashcards it is not too difficult to catch up. With lots of flashcards and a backlog of 2 or 3 weeks it is impossible. Therefore I think it is important to accept users who want to review flashcards on a more or less regular basis, but who do not care - because they cannot care - if there is a backlog. I think it is no solution to eliminate some cards from further reviews: if you write a test at school or university and fail because just the flashcards you needed were eliminated by Pleco, you could not convince your teacher.

So let me repeat the solution I use for my considerable backlog now for almost one year: SRS with an upper score limit as it was first described by Gato. That also temporarily eliminates flashcards from further testing - all cards with a score above the actual upper limit. But the upper score limit can be increased every day, and the eliminated cards are thereby reduced every day. There are many cards I still remember though they are 20 or 100 days overdue, because I spent time to learn them in the past, and SRS is not such an exact system that it always finds the right time for a review. And if I fail with a card, I do not need to learn the card completely from the beginning, but easily remember the card in the future. If you would remove a whole category from further testing, you will certainly forget most of those cards, at any rate those cards with a low score which are not yet burnt in your memory.

This system works with the actual Pleco system, but it is too difficult to use for normal users. There are many ways to support the system: Let the user determine the number of cards he is willing to test, and let Pleco set the upper score limit to reach this number of cards - that would be one way of easy support. If you fear the user does not find out that his backlog is still increasing, let Pleco tell him for instance the dayly development of the upper score limit and the backlog.

The important point is to accept and support a SRS with a backlog as one possible system besides the standard SRS with the dayly review of all due cards - the latter I use for new cards.

 

[mikelove]

Honestly, for 'normal users' even the idea of managing a backlog feels like a bit of a stretch - if we set the threshold for 'things which everyone needs to customize' low enough to include that, we'll end up with still too many options.

The ideal interface would let people choose which categories they want to study now and how many cards they want to study now and have us manage everything else related to card selection and scoring, even backlogs. (which as I've said we may address in part by removing cards from active study / adding them again when you have time) And that's pretty much our goal - the market for flashcard study systems with fine-grained algorithm control is quite limited and most of those people are probably already using Anki anyway, we want a flashcard system that *everybody* can use.

 

[gato]

I think you target just a little easier than Anki. Pleco's advantage is the integration of the dictionary and reader with the flashcards. You'll have users sticking with you as long as your flashcards is not significantly harder to use OR have less features than Anki. If you have a flashcards that have much less features than Anki, then it might force a lot of people to stick to Anki or move to Anki.

Anki has implemented a version of backlog management in Anki 2, by the way.

The question is who is "everybody"? If someone who doesn't want to spend any time learning how to use a flashcards, I question whether that person would even bother spending 30 minutes everyday with flashcards, or more relevant, bother with purchasing the flashcards.

The ideal, I am sure you would agree, would be to have a system much easier to use than Anki and the current Pleco, but with just as many features. To the extent that extra features make the system harder to use, you need to make cuts in features. But if you tried to maximize ease of use, without regard for features, you could end up with a system that is very easy to use, but with a very limited set of features. That's not what most people would want in an upgrade of the flashcard system. I don't know if that's your plan, but your recent messages about changing the system so that "everybody" can use it makes me worried that you are going to go too far in the other extreme.

 

[mikelove]

Features are not at all the same thing as options; we could make our algorithm much more sophisticated and still get rid of 80% of our current options. Backlog management is an excellent example. We can pretty easily detect that somebody has a large backlog; the question is whether we should automatically deal with that in an appropriate way, or whether we should present the user with a bunch of different buttons / options / etc to let them decide how to deal with it themselves. "Learning how to use flashcards" shouldn't even be a thing, really; it isn't for paper cards.

So the goal for us is to have more features but fewer options; ideally, hardly any of them in the card selection / scoring domains at least. We'd like our SRS to be more intelligent than it is now, but to apply that intelligence automatically without requiring the user to tinker with it. And frankly, we think we'll gain more users than we'll lose through this, and we'll also end up with something that's easier for us to maintain going forward. This was the same philosophy we applied to other parts of Pleco in 3.0, and while we did indeed get some angry emails about things we'd removed / changed, our sales and downloads are both at record highs and the vast majority of our users seem very happy with it. I could well imagine a few longtime PlecoForums users posting angry messages in 6 months accusing us of ruining our app and announcing that they're abandoning us for another flashcard system, but that doesn't necessarily mean we're making the wrong move here.

We're not living in the Palm/WM era anymore - our audience is far broader and their expectations in terms of simplicity and design refinement far greater; the flashcard system is the last remaining piece of antediluvian Pleco and it's time we bring it into the modern mobile age.

 

[sobriaebritas]

If someone who doesn't want to spend any time learning how to use a flashcards, I question whether that person would even bother spending 30 minutes everyday with flashcards, or more relevant, bother with purchasing the flashcards.

I don´t mind spending time learning how to use the flashcards, as long as I don´t have to spend more time on that than on actually learning what I mean to learn with them.

 

[HW60]

Features are not at all the same thing as options; we could make our algorithm much more sophisticated and still get rid of 80% of our current options. Backlog management is an excellent example. We can pretty easily detect that somebody has a large backlog; the question is whether we should automatically deal with that in an appropriate way, or whether we should present the user with a bunch of different buttons / options / etc to let them decide how to deal with it themselves.

I fully agree! But I think it could be helpful for the final result of the flashcard system if "dealing with that in an appropriate way" could be discussed in the forum before. I did not ask for options in my last post, I asked for a solution of the backlog problem. Certainly Pleco is able to find out that there is a backlog problem if with 5000 flashcards 2000 are due for review. But the question is how to deal with that. I doubt that "removing cards from the deck" in a non transparent way is something the "normal user" expects. Removing options is not the same as removing transparancy. But it seems that it is still too early to discuss these ideas.

 

[mikelove]

"Removing cards from the deck" isn't what I'm talking about - I'd describe it more as letting some cards get *way* behind in order to keep most other cards on schedule. Given the choice between reviewing every card very late, or reviewing some cards on time and some other cards very very late, I think the latter strategy makes more sense - reduces the number of cards you're likely to forget.

 

[copumpkin]

I've skimmed most of this thread (so apologies if this was proposed and I missed it) and was wondering about a possible transition plan from the old to the new flashcard system.

Pleco already has all sorts of optional downloadable features. I don't know whether the current architecture permits this, but why not lift the current flashcard codebase wholesale out into an optional "Legacy Flashcard System (deprecated)" downloadable module for those who depend on it? This would work like (code) library deprecation cycles: people can try the new stuff, keep using the old stuff with an eye on how to migrate it over, and then in a release or two (after feedback from legacy users) you can kill the deprecated version and everyone will be happy. The new system would be the default one that new users would get, of course.

Even if it wouldn't be a "wholesale legacy codebase resection", architecting the flashcard code to be modular in that fashion could allow for future customization with different flashcard systems (not that I can realistically think of that much space for variation in that space ) or generally help maintainability of a key feature.

 

[mikelove]

Not really possible to make that modular, unfortunately - among other things, it would be an absolute nightmare to get the old system working with the new one's database format.

But there'll be beta-testing for the big stuff and we may actually even launch some of the more controversial changes on Android first, since unlike on iOS it's relatively easy for Android users to roll back to the old app if they aren't happy with the new one.