Flashcards Simplified

gato

状元
I posted this earlier on another thread. I thought it might be useful for some as a separate thread.

-----
Suppose you have a long flashcard list and want to review 50 cards at a time until you memorize them, here is how to do it.

Go to the Card Setting for flashcard (p. 53 of manual).
- Here you can set "Sort by" to "creation date - earlier". This will fix the problem of cards showing up in random order.
- Set Mode to "Fixed # of Cards" and "50" (or however many cards you want to review).
- Set Limit Ranks to "1 through 1" (I use rank 1 to indicate cards I have not memorized, yet).

Then go to Rank Setting
- Set "Advance to next rank after" to "3 correct in a row." This will require you get the card right three times in a row before its rank advances by 1. In other words, when a card goes from rank 1 to 2, you will have gotten it right three times in a row.
- Set "Retreat' to "1" so that a card will go down a rank if you get it wrong once.

That's it. Ignore the rest of the settings for Card, Rank, and Flag.

-----
If you want to be tested on Chinese writing, you can change the setting so you are prompted with the definition and pinyin, and you have to supply the characters.

See
http://www.chinese-forums.com/showthrea ... =plecodict
Open Plecodict
Start new flashcard session
Test Settings
Change the Test and Show settings so that under Test only Defn + Pinyin is ticked

-------
I think these setting will satisfy most people. The other settings involving repetition spacing and frequency adjustment are difficult to use correctly. In any case, using the method described above to study a relatively small set of cards over and over again until you memorize them is the probably most efficient way to learn.
 

koreth

榜眼
I don't agree that that's the most efficient way to learn, because if you limit the quizzing to rank 1, it assumes that you either don't know a word at all or that you've completely memorized it after three successful answers. I will often get a word right three days in a row after learning it, then draw a complete blank on it a week later. I think it's very common to forget new vocabulary several times before it finally sticks, and for me at least, on average each time a new word sticks in my mind, it lasts a bit longer until I've finally got it. Which is exactly the process repetition spacing is designed for.

In my opinion setting up repetition spacing is definitely worth the learning curve (though if you want to just use the default spacing settings, that's dead simple and is probably good enough in many cases.) I am memorizing characters much faster than other people I know who are studying Chinese, and I attribute it in large part to the fact that my flashcard system is smart about what to drill me on most often.
 

lmcjipo

榜眼
koreth said:
I don't agree that that's the most efficient way to learn, because if you limit the quizzing to rank 1, it assumes that you either don't know a word at all or that you've completely memorized it after three successful answers. I will often get a word right three days in a row after learning it, then draw a complete blank on it a week later. I think it's very common to forget new vocabulary several times before it finally sticks, and for me at least, on average each time a new word sticks in my mind, it lasts a bit longer until I've finally got it. Which is exactly the process repetition spacing is designed for.

In my opinion setting up repetition spacing is definitely worth the learning curve (though if you want to just use the default spacing settings, that's dead simple and is probably good enough in many cases.) I am memorizing characters much faster than other people I know who are studying Chinese, and I attribute it in large part to the fact that my flashcard system is smart about what to drill me on most often.
I use a similar scheme to gato except that I use 5 correct answers in a row and every few weeks, I increase the ranks of flashcards that I'm tested on by 1.

For example, my current flashcards are divided into lessons from my textbook. I test myself on the first lesson (about 20 characters/words) and if I get the character/word correct 5 times in a row, the rank goes up to 2 (and it doesn't appear in my current flashcard test session). After all rank 1 characters go to rank 2 or a significant portion of them move to rank 2, I will set my flashcard session to test me from rank 1 to rank 2 and add the second lesson characters/words to my flashcard session (and these second lesson characters/words start at rank 1). I continue to do the same thing so that when I add the lesson 3 words, I'm being tested on ranks 1 to 3 so that I have to get the new lesson words correct more often in order for them to advance above my upper limit testing rank. I'm not sure if this makes any sense.

I agree with you that sometimes I can get the characters correct in 5 times in a row and then draw a blank a few days/weeks later after not practicing.

I'm intrigued by your use of repetition spacing because I want a "smarter" flashcard system but don't really know how it works. Information on repetition spacing in the manual is somewhat confusing and scarce.

Can someone explain it?
 

koreth

榜眼
I can explain it, I think. I will start with the general theory and then describe PlecoDict's version.

The basic theory at a very high level is actually dead simple. The more often you get a card right, the less often you see it. The more often you get it wrong, the more often you see it.

What this means in more practical terms is that a card takes a long, long time to completely disappear from your tests but it gets very infrequent after a while as you continue to get it right. Every day you review a mix of cards you haven't gotten right very much (either because they're new or because you've gotten them wrong) and ones you've gotten right a lot but haven't seen in a while.

For example, I'm in the middle of my daily quiz right now and I'm looking at a card I added on December 13 of last year, have gotten right each time it's come up, and most recently reviewed a month ago. It will be another month before I see it again, and after that it'll be a month and a half. I mark that one (correct again). The next card I'm shown was added even longer ago, on October 16, but I've gotten it wrong a few times so it's currently coming up every 3 weeks instead of every month. If I get it wrong it'll start coming up every week, then every couple days, then every day until I start getting it right.

So, how does this work in PlecoDict? Ranks are the key. A card moves up and down in ranks as you get it right or wrong. A card's rank determines how long PlecoDict waits between reviews. The trick is that you have to be using the "custom" repetition spacing mode (in Card Settings, select Mode: Repetition spacing, then select Spacing (days): Custom, then tap the Setup button). In "fixed" mode (the default) ranks are ignored. Don't use "fixed" mode.

The length of time between reviews is determined by the number of days you've entered for the rank in question. A spacing of 1 means review the card every day, 2 means every other day, etc. One special case is a spacing of 0, which means review the card an infinite number of times until you get it right enough times to move it to the next rank. Some people use that as their rank 1 spacing so they can keep reviewing new cards on the first day.

My setup? I want a smooth progression of time between reviews, so I have a lot of ranks -- 13 -- and fairly aggressive movement: 2 correct in a row goes up a rank and 2 incorrect in a rank moves down 3 ranks. My spacing values are 1, 2, 4, 7, 11, 15, 21, 30, 45, 68, 94, 141, and 211 days. By the time a card gets to rank 13 I only see it once or twice a year.

One thing about repetition spacing mode that you may or may not like is that you'll get a different number of cards to review each day, since there is a certain set of cards whose ranks and last review dates say they are scheduled to be reviewed on a given date. If you've gotten lots of cards right and wrong in the right combination, a whole bunch of them could end up landing on the same day. I usually deal with big spikes in the card count by just not finishing the list, but getting as far into it as I have time for; I catch up after a day or two.

If you want a fixed number of cards to review each day, another option is frequency adjust mode, which I don't have as much experience with but I think I understand. Frequency adjust mode is a little more fuzzy than repetition spacing mode: it basically deals in terms of, "how likely is it that card X will be shown to you next?" If card X has a frequency twice as high as card Y, you'll see card X twice as often. But it's randomized to some degree so you don't know what order the cards will come in or which specific cards it'll pick. You can set PlecoDict to review a fixed number of cards each session in frequency adjust mode, and it picks them out based on their ranks.

Though I don't use it too much, I set the frequencies to match my repetition intervals: 999, 500, 249, 142, 90, 67, 48, 33, 22, 15, 10, 7, 5. (Those are 999 divided by the number of repetition-spacing days for each rank.) In theory this should give me roughly the same spacing of cards in either mode, but in practice the random aspect of frequency-adjust mode is a little annoying since it'll sometimes give me the same high-ranked card two days in a row and not give me a particular low-ranked card at all.

You can of course switch back and forth between frequency adjust and repetition spacing; it will keep track of everything as you'd expect.

One other thing I do that I think makes testing more effective: I vary what I'm being tested on as cards move up in rank. In rank 1 I'm shown the Pinyin and definition and have to write the word. In rank 2 I'm shown the characters and have to say the word and the definition. I alternate those except for ranks 6, 8, and 11, where I'm just shown the definition and have to write the characters and say the word. That way I have at least some ability to come at a word from any angle (hearing it, seeing it, or needing to say it myself when I want to express something.) You can adjust that in "Test Settings" by setting "Display fields: By Rank" and hitting the Setup button. The "Test" line for each rank is what you're shown when you're being quizzed, and the "Show" line is what you're shown when you reveal the card. I have all my "Show" lines set to show everything.

Hope that helps.
 

koreth

榜眼
By the way, should this thread maybe be in the main PlecoDict forum instead? I thought the Flashcard Exchange forum was for announcing flashcard lists people wanted to make available. Lots of people who don't care about user-supplied flashcard lists might want to follow this.
 

mikelove

皇帝
Staff member
Moved. (sorry about that, lazy forum administrator syndrome)

And thanks very much for posting this great explanation of the flashcard system! Much better than the one in the instruction manual.
 

koreth

榜眼
Sure thing. You're welcome to use what I wrote in any way you like if some of it would help clarify the manual.
 

gato

状元
Each method has its merits, but the method I recommend is perhaps simpler and more predictable (i.e. you can control which wrods are tested more easily).

I am memorizing characters much faster than other people I know who are studying Chinese,
Is your control group using my method? :lol:
By the way, should this thread maybe be in the main PlecoDict forum instead? I thought the Flashcard Exchange forum was for announcing flashcard lists people wanted to make available. Lots of people who don't care about user-supplied flashcard lists might want to follow this.
The problem with the PlecoDict forum is that important threads gets lost; whereas the flashcards forum doesn't yet have so many threads.
 

koreth

榜眼
Of course each person should do what works best for them; different people learn differently. But
(i.e. you can control which wrods are tested more easily).
My beef with the simple method is not that you can control which words are tested, but that you pretty much have to manually control which words are tested, especially when it comes time to review (unless you're one of those lucky souls with a photographic memory who never needs to review once you learn something initially -- in which case your method is indeed optimal.)

It seems to me that your technique is equivalent to having two stacks of physical flashcards, a "to learn" stack and a "learned" stack. Each day you go through the top N cards on the first stack, and if you've gotten a card right for the third time, you put it on the "learned" stack, where it sits indistinguishable from the hundreds of other cards you put in that stack six months earlier. That's fine for hitting lots of new material, but you have no good way to go through the "learned" stack and figure out what you might have forgotten over time short of scanning through the whole stack.

If I am misunderstanding your scheme, please enlighten me -- how do you avoid that problem?

Manually choosing cards to review is totally manageable when you have a couple hundred flashcards since you can just scroll through the list and pick out the ones you want. When you have a couple thousand, all sitting at rank 2, how do you choose which ones to review later? You can do it by date, of course, but then you'll spend time reviewing lots of cards you already know well and don't actually need to review. I just don't see how you will avoid wasting way more time and effort at that point than it would have taken to set up repetition spacing in the first place. Which really only takes a couple minutes and only needs to be done once.

My feeling is, keeping accurate track of which cards I'm getting right and which ones I'm getting wrong over time is exactly the sort of thing a computer is good at. I don't want to have to control which cards to test; with repetition spacing I just hit the "start test" button and I get exactly the set of cards I need to be looking at today.

I am not, by the way, trying to claim that PlecoDict's repetition spacing is the ultimate, 100% optimal way to learn. There are a few things I wish it did differently. But I think it's better than managing cards by hand.
 

gato

状元
koreth said:
My beef with the simple method is not that you can control which words are tested, but that you pretty much have to manually control which words are tested, especially when it comes time to review (unless you're one of those lucky souls with a photographic memory who never needs to review once you learn something initially -- in which case your method is indeed optimal.)
It's not "reviewed once." You have to get the character/word right 'n' times in a row, where you set the 'n'. Getting it correct that many is a pretty good indicator that you already know it.
It seems to me that your technique is equivalent to having two stacks of physical flashcards, a "to learn" stack and a "learned" stack. Each day you go through the top N cards on the first stack, and if you've gotten a card right for the third time, you put it on the "learned" stack, where it sits indistinguishable from the hundreds of other cards you put in that stack six months earlier.
It's more than just 2 stacks of cards, but rather 12, or however many ranks Pleco has.
Manually choosing cards to review is totally manageable when you have a couple hundred flashcards since you can just scroll through the list and pick out the ones you want. When you have a couple thousand, all sitting at rank 2, how do you choose which ones to review later?
The flashcard list I use the most right now is the one I uploaded in the other thread of 2200 characters. If you can get a character correct on three times in a row on separate days, you know the character very well and should be able to move on to another characters. The key here might be what does it mean to get it "correct." You have to set the definition yourself and be honest with yourself when you do the review. I think at the minimum it should be recognizing the character, knowing the pinyin and the definition. I also keep a separate flashcard list to test myself on being able to write a character when given a pinyin+definition prompt. If getting it three times in a row is not working for you, then maybe you can set it to five times in a row like one of these earlier posters said.

As for how to review cards that in a higher rank than the one you are currently reviewing, you can simply change the setting to "limit to rank n to n" where 'n' is the rank you want to review. When reviewing the cards in rank 'n', if a particular card is easier for you than the rest in that rank, you can manually bump up its rank with the edit option. Then you won't see it again the next time you're reviewing that rank.

One thing about repetition spacing mode that you may or may not like is that you'll get a different number of cards to review each day, since there is a certain set of cards whose ranks and last review dates say they are scheduled to be reviewed on a given date. I usually deal with big spikes in the card count by just not finishing the list, but getting as far into it as I have time for; I catch up after a day or two.
How many cards would you have to review if you have a flashcard list of over 2000 characters, which I do? 200 to 300? Do you have to review all 2000 on the first day? It takes me over an hour to review 100 characters; I don't think it's practicable to review more than 100 per day. Doesn't it defeat the purpose of the method if you just not finish the reviewing? This is the biggest problem I see with repetition spacing.
 

koreth

榜眼
I'll digest most of that when I'm more awake tomorrow, but as for the last bit (how to deal with a ton of cards), it's not actually a problem. Yes, you will be presented with a 2000-card quiz the first day. But the spacing is based on the number of days since you last reviewed a card, not the number of days since you were last scheduled to review it. So any cards from day 1 that you review on day 2 will use day 2 as the basis for figuring out the subsequent review days.

You would want to sort the cards by one of the non-random sort orders, probably "reviewed (latest)", so you'd get the same cards at the start of each day's test until they advanced in rank enough to not appear every day. If you used random sorting you'd get a different subset of your 2000 cards every day which wouldn't be too useful. If you use "reviewed (latest)" sorting you will get, in order,

* the rank 1 cards you've already seen (the "haven't gotten them right 3 times yet" list)
* the higher-ranked cards you've already seen, in order of increasing rank (since the higher the rank, the further back the last reviewed date)
* all the cards you haven't seen

With that in mind, you just do as many cards as you have time for each day, and after a short while, as you advance those cards' ranks, the total card count for the day will go down. You would probably want to stop after seeing a certain (relatively small) number of new cards so as not to hit yourself with a bunch of new material at once.

That said, I have never started with a huge flashcard list on day one. I have been building up my flashcard lists a little at a time, adding new words each day as I study. If I were working from a big preexisting list of material that was brand new to me, I would probably be inclined to start a second list and move cards from the first list a few at a time so as to let me see in advance which words to go study offline before I start quizzing myself. Moving cards off the main list would make that easier because, as in your system, you'd always know that your new words for the day are the first N cards on the big list. I'd quiz myself from the second list and there would never be an issue with having a huge card count in test mode.

Working with big preexisting lists is probably something that the revamped flashcard system in Pleco 2.0 can improve on. I agree it's a little awkward right now no matter how you slice it.
 

lmcjipo

榜眼
koreth said:
I can explain it, I think. I will start with the general theory and then describe PlecoDict's version.
...
Hope that helps.
Thanks. Your explanation helped me a bit. I am now trying the repetition spacing feature of the flashcards.

However, I don't understand the following screen.

In my "Card Settings", I select "Repetition spacing" for Mode and "custom" for Spacing (days). When I click on "Setup" (to define my custom spacing), I'm assuming that the numbers on the far left is the rank. I understand the "Spacing" field but I don't understand the "Freq:" (frequency) field.

If for example, I have:
1 Freq: 100 Spacing 0 days
2 Freq: 90 Spacing 1 days
3 Freq: 80 Spacing 2 days

I'm assuming that cards that are of rank 1 will be shown every single time that I initiate a flashcard session (which is what I want). Cards of rank 2 will be shown every 1 day but what does frequency mean? Does this mean that only 90% of the cards that have a rank of 2 will be shown every 1 day or does it mean something else?

Thanks.
 

koreth

榜眼
The frequency numbers are ignored when you're testing in repetition spacing mode. They are for controlling which cards appear more often when you're doing a test in "endless loop" or "fixed # of cards" mode. Basically if one rank has a frequency twice as high as another rank's, the first rank's cards will tend to appear twice as often.

If you are only planning to use repetition spacing, you can leave the frequency numbers alone.
 

redpixel

秀才
koreth and gato thanks to both of you for posting this!

We currently have exam periods over here at BLCU so I've been inclined to use the gato method for just cramming as much as I can :)

From now on I think I'll try to add flashcards whenever I see an interesting word though and use the koreth method for reviewing a few cards each day. I'd like to know, do you guys usually split things into several lists or just keep it as one big list? Provided that you're starting from scratch and adding whichever words you come across.

And yes, this explanation ought to be put as a sticky somewhere so it's easy to find :)
 

gato

状元
As you may already know, you can designate a list to be the "default list," which means that an future addition to flashcards (using the big PLUS icon) will be added to that list.

I prefer that my "completely new vocab list" (as opposed to 2200-character list that I'm using to practice writing) be limited to 400 words or so. So I would create a new default list whenever an existing list gets too big. A smaller list gives me more flexibility in choosing what to review. I name these flashcard lists with the month and year when they were created. I usually start a new list every two or three months.

I spend most of my time reading books and magazines in Chinese. I'm at a level where there may be 10 new vocabs or so in a typical 2-page magazine article. I usually would underline them and guess what they by context, rather than looking them up right away. I would look up the new vocabs all at the same time only after I'm done with an article, and sometimes only after I'm done with an entire magazine. This works when there aren't too many new vocabs and help to cut down the lookup time. Every word I look up I add to the default flashcard list, which I try to review regularly. I think if you follow this approach, you'll see that you'll have to look up fewer and fewer words.
 

longjie

Member
Flash cards need simplification

PlecoDict rocks generally. It's probably the single most useful product I have for learning Chinese.

However, flashcards are a disappointing feature. The settings are overly complex, and the complexity does not add value. E.g., who cares whether I see a 6 ranked card in 110 or 130 days?

Yet even with this complexity, it's damn hard if not possible to do the basic thing I want:

1) Have lots of cards (maybe 500)
2) Learn them progressively in chunks of perhaps 10, with the review in randomized order, but the cards the SAME until I learn them
3) Occasionally review learned cards and reintroduce forgotten ones
4) Be able to study as much as I want in when I have time (the repetition spacing features is terrible for this - I want to study when I have time - not when some essentially arbitrary random noise pattern happens to hit a high point that day)

I still have not figured out how to do this. Even if it is possible, it's WAY too difficult, given that this would suit almost everyone's needs. The current settings are essentially an inflexible scripting language, which may be great on occasion, but certainly does not suit most of who don't have time to mess around with all these settings. I worked as a software engineer for 6 years and I can't quite figure the thing out.

Basically the settings need to be much much higher level, like "cards to review at a time", "learning sensitivity (i.e. how quickly cards become less frequent when they are correctly answered)" ... Actually I'm not sure you need any more than that.

One other feature I'd like to see is the ability to crossreference characters without leaving the review. Often I get a new word and want to know what the individual characters mean while I'm learning.
 

mikelove

皇帝
Staff member
I completely agree with you about the system's complexity (as I've said several times in this forum). Though if you think that's bad you should have seen the first version, which really was a scripting language and let you configure the exact conditions (with comparision statements, AND/OR clauses, etc) under which a card would be promoted to a particular rank. The issues with the current system are a big part of why we're giving away the (non-stroke-order-or-audio-dependent parts of the) flashcard system in 2.0 away for free.

However, your description of your ideal flashcard system illustrates nicely why it's difficult to make things as simple as we'd like to. There are many users who would strongly disagree with you on requirement #4 - we're constantly getting requests to make our software more like a popular flashcard program called SuperMemo, which, much like our repetition spacing feature, only presents cards for review that it thinks you need to review on a particular day based on a similar "arbitrary random noise pattern." So already that's another checkbox to add to your two settings. We've also got people who want to review cards in the same way that they'd review printed ones - pick a list of cards, review all the cards in it in random order, and manually decide when they had learned a card well enough to drop it from the list. Which necessitates at least one more checkbox. If you browse through the forums you'll see a great many other examples of people's study patterns that would necessitate an additional option or two.

I agree with you that obscure settings like the # of ranks or the exact # of days between repetitions are rather silly, though, and it's likely that those will indeed disappear in 2.0. Flags are already being eliminated in favor of allowing cards to be assigned to more than one list at a time, so it's likely that in 2.0 the Rank/Flag/Card settings screens will be replaced with just one screen (and no separate "custom" screen) - not quite the minimalist ideal you describe, but a lot less confusing than what we have now.

Cross-referencing will almost certainly be included in 2.0 - it's another thing we've gotten a ton of requests for, and since we're already planning to include it in dictionary definitions it's trivial to extend it to flashcards as well.

Thanks for the feedback!
 
Koreth flashcard system is really great! I just love that!

It really takes some deep thinking to understand the finer points. One point I noted:

My setup? I want a smooth progression of time between reviews, so I have a lot of ranks -- 13 -- and fairly aggressive movement: 2 correct in a row goes up a rank and 2 incorrect in a rank moves down 3 ranks. My spacing values are 1, 2, 4, 7, 11, 15, 21, 30, 45, 68, 94, 141, and 211 days. By the time a card gets to rank 13 I only see it once or twice a year.
I think (and please correct me if wrong) when you start at rank #1 and spacing values are 1 then you can test only once a day. Right?

I did one test with that setting and got an error when I wanted to have a 2nd try. So I set spacing values are 0 for the first entry. Then the test could start and only those previously wrong once popped up.

So my setting is now: (sorry, no screen shots)

Test (not complete, only relevant stuff)
self scored
fixed
rank list

Cards
repetition spacing
custom
random
(rest is un-checkd)

Cards >> Setup (
freq spacing
100 - 0
66 - 2
44 - 4
29 - 7
19 - 11
12 - 15
8 - 21
5 - 30
3 - 45
2 - 68
1 - 94
1 - 141
1 - 211

Ranks
13
2 correct in a row
retreat 3

Flaks
nothing here....

Any suggestions thereafter?

I think a Pleco人 should do screen shots with some common setups for different strategies. I think it's great that the cards have endless possibilities, but that makes them very hard to setup.
 

ipsi

状元
1) This thread (or parts of it) should be stickied :).

2) I've had liked a scripting language... But I'm strange like that.
 
All this talks about repetition spacing and ranks reminds why I stopped using the pleco dic's flash card program. I've been using Pleco dic since it came out, works great on my treo. Yea, the repetition spacing is great when taking a course and you're studying every day, but when your working, traveling, or doing anything else in which you might take a week, month or year hiatus from studying regularly, getting back into those old decks of flash cards on pleco dic is unmanageable. Which is why I still stick to my nuts and bolts Flash! program after all these years. But please correct me if I'm wrong, I didn't use the flash function for very long.

For me it comes down to one little thing: you can't organize the decks by meaningful categories (like topics, or chapters in a book) which makes it hard to just have a quick review of only the things you want. So with Flash! I've got about 30 decks, most of which are for books of Chinese I've studied (and some other topics, like 成语s food words etc) and each deck is ordered by chapter for an estimated 8000 words. So if I want to quickly review a book, I can pick the appropriate deck, and then go through chapter by chapter testing myself on the words. Then when I have the whole book done, I can review myself on the whole deck at once. Having 8,000 words in Pleco dic and trying to organize them would just be a mess since you can't divide up the decks into categories.

I definitely see the benefits of repetition spacing and wish I could combine it with the my old method, but since I study irregularly I just stick with this nuts and bolts program (copying and pasting the definitions from pleco dic of course). I hope Pleco 2 finds a balance of both worlds.

Again, please call me out on this or teach me how to use the Pleco dic better, since it's flashcard function is superior for the most part. Or if anyone knows of any other good flashcard programs for the palm, please let me know so I have something new to hold me over till' Pleco 2. My current program is circa 1998.
 
Top