Card Create "+" key seems to violate documented format

[Jim Kay]

I've created a few cards using the "+" key and am programming according to the documented format. But these cards seem to vary in their adherence to the documentation. If import them as flashcards and take a 'test' the experience is rather odd and unpredictable.

Most of the cards show "simplified"["traditional"] even when both forms are the same. "fang2" does not and "jiang4" does not.

"fang2" has a boat load of additional definitions. (A nice feature, but the format is undocumented and the resulting flashcard test treats the additional definitions as some kind of incomplete/invalid cards.

What is going on here?

直觉[直覺]zhi2jue2(NOUN) intuition
柚子[柚子]you4zi5(NOUN) pomelo
躲闪[躲閃]duo3shan3to evade/to dodge (out of the way)
房fang2◆
① house 一所磚∼ Yi1 suo3 zhuan1 ∼ a brick house
② room; bedroom 三間∼ san1 jian1 ∼ three rooms 兩∼一廳 liang3 ∼ yi1 ting1 two bedrooms and a living room
③ a house-like structure 蓮∼ lian2∼ lotus pod
④ a branch of an extended family 三∼ san1 ∼ the third branch, i.e. the third son and his family
⑤ {measure word} (for daughters-in-law or concubines) 有兩∼兒媳婦 you3 liang3 ∼ er2xi2fu5 have two daughters-in-law 娶一∼小老婆 qu3 yi1 ∼ xiao3lao3po5 take a concubine
⑥ the fourth of the twenty-eight constellations (二十八宿) into which the celestial sphere was divided in ancient Chinese astronomy (consisting of four stars nearly in a straight line in Scorpio)
⑦ (Fáng) a surname

◆ same as 坊 fáng
匠jiang4craftsman; artisan 鐵匠 tie3jiang4 石匠 shi2jiang4
资讯[資訊]zi1xun4information
详细[詳細]xiang2xi4detailed; minute ∼的報導 ∼ de5 bao4dao3 a detailed news report ∼了解情況 ∼ liao3jie3 qing2kuang4 acquire detailed knowledge of the situation ∼占有材料 ∼ zhan4you3 cai2liao4 collect all the available material; have all the relevant data at one's fingertips ∼地描述 ∼ de5 miao2shu4 give a minute description 請說∼點。 Qing3 shuo1 ∼ dian3. Please explain in greater detail.
资讯科技[資訊科技]zi1xun4ke1ji4• information technology • science of communications
龙[龍]long2• dragon M: tiao2 [條] • imperial
半价[半價]ban4jia4• half price

 

[mikelove]

Did you use the "word list" feature to create these cards? Honestly, that feature is about to be removed and we haven't done much with it in years - it was a temporary feature we added in the few months after we released our iOS app before we'd ported our flashcard module to it - so I'd recommend disabling it.

The simplified[traditional] business might have to do with cards that have custom definitions, or cards that were formatted that way on import - would either of those apply in this case?

 

[Jim Kay]

I was happily looking through the dictionaries by typing or writing Chinese. When I found something interesting, I'd add it to my card list which I could import later (or copy to a PC and look at it there.)

When I did import it later as flashcards, and browsed it (edit) or tried to take a 'simple profile' test on it, I got really strange results with those multiple-definitions.

 

[mikelove]

I was happily looking through the dictionaries by typing or writing Chinese. When I found something interesting, I'd add it to my card list which I could import later (or copy to a PC and look at it there.)

When I did import it later as flashcards, and browsed it (edit) or tried to take a 'simple profile' test on it, I got really strange results with those multiple-definitions.

Yeah, in that case I'd definitely recommend disabling that "word list" feature (in which case the + button will simply create a flashcard).

 

[Jim Kay]

And if I do that and then export it, will it be better?

 

[mikelove]

And if I do that and then export it, will it be better?

Yes, the Export command is much more reliable about stripping newlines.

 

[Jim Kay]

Certainly, newlines are a major part of the problem. But a flashcard with a vast, long list of 'translation' data is a problem as well. The program I export them to is loading the data into SQL server, currently the 'translation' is a text field and that's limited to 255 bytes. There are ways to make the field a 'blob' but virtually any study program is going to have trouble with so many definitions. Flashcards are not well suited for multiple answers.

Consider a Chinese character where one character can have just one of its pronunciations that have multiple possible meanings. How do you test giving the character where the answer is any of several translations. Or the situation where one 'translation' can be matched to multiple different characters?

It's almost impossible to study a dictionary with flashcards.

 

[mikelove]

Certainly, newlines are a major part of the problem. But a flashcard with a vast, long list of 'translation' data is a problem as well. The program I export them to is loading the data into SQL server, currently the 'translation' is a text field and that's limited to 255 bytes. There are ways to make the field a 'blob' but virtually any study program is going to have trouble with so many definitions. Flashcards are not well suited for multiple answers.

Even a single definition can easily run more than 255 bytes, so I'd recommend changing that schema.

Consider a Chinese character where one character can have just one of its pronunciations that have multiple possible meanings. How do you test giving the character where the answer is any of several translations. Or the situation where one 'translation' can be matched to multiple different characters?

The one-translation-to-many-characters problem is going to exist with pretty much any language flashcards - the only way to prevent it is to provide an additional piece of information to narrow it down, which is why we also offer the option to show the pinyin or character (with you then expected to fill in the other one).

As far as characters with lots of meanings: frankly, most textbooks / curricula these days have adopted more of a word-based approach; many of them don't even bother with single characters at all initially, except for characters like 了 which primarily function as standalones. So we haven't seen much demand for enhanced study features for single characters, and the biggest request we have seen is for mnemonics to help you remember how to write them, rather than for better support for reviewing lots of definitions.

So the ability to create multiple cards with different sub-definitions for the same character + pronunciation is not an unreasonable request, just not a very popular one - we get maybe one or two emails / forum posts a year on the subject.

It's almost impossible to study a dictionary with flashcards.

I'm sorry you feel that way - if you email support we'll be happy to set up a refund for your purchase, since it seems you would prefer a standalone flashcard system rather than one that's integrated into a dictionary.

 

[Jim Kay]

I don't actually remember what that feature cost (and the App store won't tell me because it's already installed.) I'm not sure it's worth bothering to get a refund as I might decide to use it again after you make more of the changes you've mentioned here and there. Would you be able to remind me of the cost?

 

[mikelove]

I don't actually remember what that feature cost (and the App store won't tell me because it's already installed.) I'm not sure it's worth bothering to get a refund as I might decide to use it again after you make more of the changes you've mentioned here and there. Would you be able to remind me of the cost?

Anywhere from the equivalent of US$10 to $15, depending on whether or not you used an educational discount and which currency you purchased it in.

 

[Jim Kay]

I didn't use the educational discount. It's definitely not worth bothering with a refund. (And I might not be eligible as it may be included in the 'maximum package' I chose.)

 

[mikelove]

I didn't use the educational discount. It's definitely not worth bothering with a refund. (And I might not be eligible as it may be included in the 'maximum package' I chose.)

OK, well I hope we are able to make it better for you then. Broadly speaking it seems like your two main problems are:

a) Lack of transparency in the way that repetition intervals are calculated; and
b) Lack of built-in support for creating multiple cards from a single dictionary entry with different sub-entries.

Is that correct? For a), have you played around with the "card filters" at all? They're pretty transparent - "# correct in row", etc - and while they're not quite the same as "boxes," they would at least let you review a particular subset of your cards based on their recent review history in a sensible way.

 

[Jim Kay]

Mike, I respect and appreciate your opinions.

But you cannot have the perspective of a newcomer to this system. You are intimately familiar with its ins and outs while a newcomer has only the documentation and actual trials. Even the documentation reads differently to a newcomer because the meanings are based on no background while you know exactly what is being addressed by each word. (Possibly you wrote it all.)

An example: with the "Simple" profile, there are "Max number of cards" and "Show." But select the "Spaced Repetition" profile, and both of those options vanish. Frankly, I cannot even begin to understand why I cannot use "Spaced Repetition" and still control those two settings.

And that is only one of dozens of bewildering facts. I've never seen the actual code and it continues to appear to me that features of different learning theories were merged into the previous collection of feature each at different times.

If there is an overall design, I can't see it.

I'm not an IT beginner but every new language and new system makes me a kind of beginner. Have you or Pleco ever tested a group of sophisticated people completely unfamiliar with Pleco and given them a series of Flashcard usage scenarios to complete?

I think you would all be really shocked to see how difficult and confusing your Flashcard feature actually is.

Anyone willing to simply add cards and settle for whatever those three default profiles do will be happy.

Anyone else won't be particularly happy.

And anyone like me, who wants a thorough understanding of what (s)he's doing, is almost certain to give up.

 

[mikelove]

I'm actually well aware of most of the problems you've talked about - it is indeed an intimidating system if you don't want to trust the default settings. You don't need a focus group to convince me that the system needs to be streamlined.

To be honest, though, I'm not quite sure how to make something transparent and sophisticated at the same time - the systems you've talked about are more transparent than ours, but they're also a lot simpler, and many people would compare them to other systems like Anki and Skritter and feel they were inadequate. As I've said, we started out with something very similar to the Leitner Box algorithm (see the manual for our old Palm software here for example - we had an elegant little system of "ranks" which behaved exactly like those boxes), but after years of aggressive lobbying by users / unflattering comparisons to other systems we eventually gave in and introduced the current "score"-based system. And confusing though it may seem, the general response to that change among our users was and continues to be quite positive - most people actually do seem content to trust the default settings and are satisfied with the results when they do.

So I don't really see a way to make the system more transparent without compromising its functionality, which makes me feel like the logical course is to make it less transparent instead, with fewer options and less access to its algorithmic underpinnings than there are now; that will at least make it easier to use, even if it disappoints people who'd like to know exactly how all of the math works. (we can still expose that too - let people view raw card statistics and provide a paper on our website explaining what the algorithm does with them - but it would no longer be quite so configurable) I'm not sure if that would fully satisfy your needs, but hopefully it would at least work better for you than what we have now.

However, changing anything is an uphill battle for a company that's been doing this for as long as we have - every particular quirk or seemingly unneeded option has its own small but vocal group of supporters - and on top of that we've had the ongoing struggle to stay up-to-date with iOS and Android the last few years - it's tough to make progress on anything when you're spending 2/3 of your programming time either porting to a new platform or making your software compatible with a new version of an already-supported platform. So awareness of these problems has not yielded a solution to them as yet.

But we've just about finished applying a much-needed revamp to the dictionary and document reader portions of Pleco - see the New Version Screenshot Thread for a little more on that - and assuming that we don't have to spend the rest of the year dealing with new issues introduced by Android 5 and/or iOS 7, our next main focus is the flashcard system, and we're fully prepared to cut out as much as we have to in order to make it sensible. (we're even considering the somewhat-radical step of releasing a "long-term support" legacy version of our software without the changes, receiving major bug fixes and nothing else, so that we can have a way to respond to people unhappy with the changes other than adding whatever feature they're missing back in)

An example: with the "Simple" profile, there are "Max number of cards" and "Show." But select the "Spaced Repetition" profile, and both of those options vanish. Frankly, I cannot even begin to understand why I cannot use "Spaced Repetition" and still control those two settings.

That one's actually covered by the manual:

Max number of cards sets the maximum number of cards you'd like to study in this session. Select "Endless" to display cards in a never ending loop until you manually end the session; select "All" to display every available flashcard (i.e., every flashcard that matches your Categories / Card Filter / etc settings) exactly once. This option is hidden in repetition-spaced tests, since in order for those to work correctly you need to study all of the cards that are currently due to be tested, though you can exit a repetition-spaced test early (or save it to resume later) if you run out of time.

 

[Jim Kay]

I actually know what the manual says on this score. But that blanket and obscure justification isi uniformative at best. I don't care how repetition-spaced wants to work, I just don't want to see more than 30 cards in one sitting. The manual gives me no clue about why I can't have that.

 

[mikelove]

I actually know what the manual says on this score. But that blanket and obscure justification isi uniformative at best. I don't care how repetition-spaced wants to work, I just don't want to see more than 30 cards in one sitting. The manual gives me no clue about why I can't have that.

This particular limitation would apply even with a Leitner Box system - if you're supposed to study a particular box after 7 days, and you stop after you've only reviewed half of the cards in the box, and the next day you only review half of the cards due that day, you're not going to review cards when you're supposed to and you're going to start falling behind / forgetting them.

 

[Jim Kay]

Hey, this is really interesting! When I view your message in Outlook, there is another line of text that is invisible when I go to respond! This is what I see in Outlook:

But when I go to reply, all I see is:

As to that disappearing first sentence: what's obscure is understanding what is due to be tested and when and why. If I were to browse through my card list and ask myself if any particular card would appear were I were to begin a test, I would have absolutely no idea. I might as well do random guessing. I think that qualifies as 'obscure.'

I tried everything I could think of to 'uncover' that missing sentence because I wanted to use your wording. But Google Chrome, at least. would not show it to me.

 

[mikelove]

Sorry, that was in a draft copy which I subsequently deleted - not sure why it ended up in the email notification, have to check on that. (or perhaps write final drafts in a different editor before posting them here)

That line was meant to justify why we don't make it optional, not to explain how the intervals are calculated; we don't make it an option because if we did, you wouldn't study all of the cards you needed to study and you'd fall behind schedule. As I said, this is a principle that would apply with pretty much any system of scheduled review intervals, including Leitner.

But to clarify, the "why" of why a card is supposed to be tested is quite simple:

last reviewed date + interval >= current date

The two dates should be obvious - "interval" in our system is:

("score" / 100) days

It's only when you drill down to "score" and how that's calculated that things get complicated, but with the "manual" scoring options you even have fine-grained control over that.

 

[Jim Kay]

But I don't have a schedule. I just work at my own pace and I decide an ' interval' has passed whenever I feel like deciding that. I'll be memorizing Chinese for the rest of my life, no hurry about that.

All I want it some unspecified unit I'll call an INTERVAL. Each time in-a-row I get a card right (be it look at Chinese and remember English or the opposite--pronunciation makes a nice item to reveal on demand) I don't see it again for an increasing number of intervals. After some specific number of sequential correct scores, it gets archived so I don't study it any more. (But I'd like to know it's still there so I can view it or unarchive it.)

1. So I decide when an INTERVAL has passed. I can do five INTERVALS in a day or just one in two or three weeks.
2. I never have to look at more than some specified number of cards during a single study session. (I happen to like 30.)
3. When I'm studying, I want to choose English to Chinese or Chinese to English. I don't want the computer mixing that up according to what IT decides.
4. This set of requirements is really only feasible if I can separately study cards according to how many times in a row I have gotten them most recently right. Otherwise the '30' number is impossible. (I have developed a sequence of going through all of those one time and then through whatever is left in the zero-right category. If the zero-right category gets too low, I pull some cards out of the Archive to repopulate it.)
5. If an eligible. set has more than 30 cards, a random selection of 30 is fine. A way of sending the excess back to the Archive would be nice too--as long as there is a way to pull things out of the archive and back into the pool to be studied.

Those are the basics of what I'm looking for. I just can't be the only person studying for personal reasons without any real schedule.

(I tried drilling down to manual scoring. It was so ' fine grained' I got completely lost in the grains and had no idea what was going to happen in a test. So I tried a test and had no idea what was happening. I guess that qualifies as an accurate prediction of consequences. But accurate isn't always useful.)

 

[mikelove]

No need to bother with scoring at all for this - did you see my earlier comment about "card filters"? Go into that screen and turn on a "History Filter" to require that a card be correct / incorrect at least a certain number of times in a row. ("not correct" and "not incorrect" basically do the opposite, make that a < instead of a >=; we ought to have a > / < / etc type list instead, but this was a way to add that feature without changing the UI)