Feature Suggestions

[mikelove]

Thanks for your thoughts on this.

The "Frequency-adjusted" card selection option would actually give you a similar effect to what you're looking for - in that mode, the likelihood of a card coming up in a test is weighted in inverse proportion to its score, so learnt cards are still appearing but less often. Another option would be to keep "Random," but with the "Unlearned card ratio" option enabled - configure the "Learned if" threshold to whatever your score cutoff is now and that'll give you a session consisting mostly of new cards but with a few old ones being mixed in each time.

As far as your suggestions, I'm not really wild about setting specific cutoffs relative to the card distribution since that's just as likely to require tweaking as your current method - someone could add a bunch of new vocabulary and suddenly find need to use a different bottom % cutoff. I'm also not big on settings which force users to juggle around specific numbers / thresholds and the like - there's already way too much of that.

So I prefer something that's a bit more intuitive - the basic idea at the moment is something which uses repetition-spaced as a baseline (guaranteeing you'll see each card at least as often as you need to to avoid forgetting it) but allows for some flexibility about what to do after you've finished all of your rep-spaced tasks.

 

[harrycallahan]

Thanks for reply. Actually I'm not surprised that it already is capable of doing something similar, given the plethora of options I figured there'd probably a set-up that would approximate what I was looking for. The problem is it's all kind of under the hood, I would like to be aware of the nuts and bolts of the implementation, same reason I've picked manual scoring. I will look into the documentation better, Frequency Adjust and Spaced Repitition are two options I haven't properly investigated. Possibly no threshold plus Frequency Adjust then will be fine, but it would be good to be aware of and have some control over specific parameters.

 

[mikelove]

Fair enough. There's nothing particularly secret about the algorithms for those options - Frequency-adjust is just a simple inversion, take a number that's larger than any possible score, divide it by the card's actual score, and use that as a weighting factor in a function that randomly draws cards from the pile, and Repetition-spaced we just divide the score by the number of points per day (100 by default).

 

[sinoreen]

hi Mike!
I'm not sure whether this suggestion has been mentioned before. When I add a Flashcard and enter a word in the "from dict"-tab and then switch to the "Custom"-tab I'd like Pleco to copy the dict entry just shown in the "from dict"-tab so that I can then add a sample sentence or other remarks. That way I wouldn't have to copy it every time manually, as I am doing it now.
I often hear a word in a sentence and then find the dict entries very good and clear, but still want to add the sentence I just heard. I found that this way it is much easier for me to remember words, because when I review the FC and then at the bottom of the dict explanation see the custom sample sentence then I immediately remember when and where and how this word was used in real life.

Another thing I'd love to see is that Pop up-Defintions also work in the Edit-Card Screen. Sometimes when I organize my FCs I look at the definitions of some of the words (via the Edit Card Screen) and then quite often want to look up words or characters in the definition.

 

[mikelove]

That has been mentioned before, but it's a licensing issue unfortunately - it's OK with some dictionaries but not others, so until we make some significant upgrades to our copy-protection system to allow separate policy settings for individual dictionaries, we have to keep that ability limited. Might also be able to get a similar thing with the new "notes" feature we're planning to eventually add to flashcards.

The way the windowing system works on the Edit Card screen makes it a bit tricky to enable popup definitions in there, but certainly should be possible eventually.

 

[thph2006]

Re: Feature Suggestions - Cursive

Hi,

It would be great to have a mode where the stroke order display draws a cursive version of the character rather than the standard print version.

Thanks, tom

 

[mikelove]

Nice, but extremely complicated - we'd have to license a whole new set of stroke order data for that, assuming that data set even exists somewhere. So this might not be feasible for a while.

 

[thph2006]

How about a mode in the stroke order screen where you can display two characters side by side for comparison? I'll often get a flashcard wrong because I confused a character with a similar looking one. When I do that I usually end up going on-line to MDBG to display the two characters' enlarged images side by side.

 

[mfcb]

hi mike,

i might have mentioned it already one time, but in the recent days i find its more and more of a problem for me.

i display the actual card score in the flashcard session. i cant prevent my eyes for using that information to interpret the headword (from the score my brain deducts the age of the card and therefore reduces the amount of possibilities). so the only option i have is to remove that information from the flashcard screen (i know how to do that, thats not the problem ). but the problem is, i still want to know, how many cards i have to learn in the different score-ranges.
its important for me, as i have to know, when its the right time to add new cards, if i have to intensify the study (at least on that day) or whether i am running out of cards to review (mostly due to clustering of cards for certain days, mentioned in some other thread).

so i just had the idea, that it should be pretty easy for you to display an information screen at start of a flashcard session, that shows statistics about the selected cards. when the session starts you know the cards involved, you also show the total amount, so i guess it should be possible to do a small table like in the statistics screen for the categories, splitting the cards into score groups and show the amounts. a simple popup to show this list would give me (and possibly others) enough information to remove the score information from the flashcard screen, hehe, i could use that position for the current time then... with a simple option to enable/disable the popup you could satisfy also the people that dont want click away another screen before they can do their session....

 

[mikelove]

thph2006 - great idea - the UI would have to be one of those dynamically-resizing-for-landscape-displays ones, though, so it'd be a trickier-than-normal dialog to implement, but certainly should be possible at some point.

mfcb - great idea also. Though it might fit into the UI more elegantly if we just added a "current session settings" option to the Statistics screen - that'd take the information from your current profile and use it to generate statistics for the cards that you would be tested on if you started a session right now.

 

[taijidan]

On the topic of stat screens: is there any way the screen shown at the end of your flashcard session could be kept in memory?
I often find I click ok at the end of the flashcard session, but then want to go back to double check something later.

 

[mikelove]

Could be, but it'd be a lot of work to add an extra "session history" table just for that - we'd really need some other features we could hang on the same data. Which we could probably come up with, but that makes this a complicated, long-term type improvement.

 

[thph2006]

Character decomposition

If at all possible please add a character decomposition feature that breaks a character into it's major sub-components and allows you to click through to definitions for each of the sub-components including the character's radical. For the radical it'd be fine if at a minimum you displayed its pinyin name, although a 1-2 word definition would be even better. The MDGB on-line dictionary is a good example of the level of detail I'd love to have for decomposition drill down.

On a side note I'd like to say I've been using Pleco for a couple weeks now, probably haven't even discovered most of the features yet but I already find it to be one of those rare programs that's amazingly natural to use. It seems to anticipate what I want to do and provides a very easy mechanism for doing it in the right place at the right time. Kudos to the Pleco team (is there anyone other than Mike?). You've obviously listened very carefully to your customers.

 

[mikelove]

Thanks! Already in the works due to a new data set we've licensed, though the interface may be rather crude at least initially.

 

[Mark Kang]

I've got my 2.0 up and running, it is definitely a generational improvement,祝贺！

I am going to echo thph2006's request- to have a function that specifically addresses characters that look similar and are easily confused.

The problem of easily confused cards/characters still plagues me after several years of Chinese language study. Consider these groupings: 浮俘淫，怨怒，竭揭褐渴,夸奈。Such groups I think cannot be made based on universal rules, although I did have a teacher once who attempted to do just that. It seems more of an individual student's brain weakness. I would like to practice these groups by themselves, but there are usually no more than 4 in each, which doesn't merit it's own category. I would like to be able to link these characters together by my grouping, then when I am doing a regular test, be able have a button that opens up this grouping in a CHARACTER SHUFFLEBOARD。The characters of the group would be laid out randomly, and tapping them would bring up the definition in a zoomed box. Tapping the box again would shrink it back down to just the character. Of course the board would have the option of reshuffling the set of cards. I wouldn't need to have any scoring or editing function for this, just the chance to mix them up and see a group of characters I easily confuse displayed in one place.

Another way to allow for this extra practice might be this- Be able to link these easily confused characters together. Rather than a special shuffleboard, provide an option that would force all the characters of any linked group to appear together in one test session. For example, if 怨 is to be tested, than Pleco would also force 怒 into the session. Although not as 好玩 as a shuffleboard would be, it would still give an extra boost of practice for a specific problem.

I would like this for my next birthday. If you can get it done in time. I will tell all my family to take the money they were going to spend on me and send it to you.

 

[sfrrr]

My solution for a similar problem (with tones for homophones) is to create a special flashcard set to drill myself. It's been working pretty well. Also, I find that remembering the meaning of the radicals often helps me remember which one goes with which character meaning. http://www.plecoforums.com/viewtopic.php?f=11&t=1845#p13689

 

[Mark Kang]

My solution for a similar problem (with tones for homophones) is to create a special flashcard set to drill myself. It's been working pretty well. Also, I find that remembering the meaning of the radicals often helps me remember which one goes with which character meaning. http://www.plecoforums.com/viewtopic.php?f=11&t=1845#p13689

Yes, I do this now, but I was hoping for a more elegant solution. If Mike takes the bait, I'll be in flashcard heaven.

 

[harrycallahan]

Hi, I see previously you're discussing Character Composition, radical sub-components. Now that you have this reference coming it's possible you have this suggestion in the works already, unfortunately with 40+ pages here it's too much to keep track of what's been suggested already.

For me this is a biggy, would be fantastic if Pleco could do what I'm about to describe.

I'll start with an analogy. Lets say we want to determine if the user knows what a tiger is. We present pictures of a chair, an apple, a book, and a tiger. The user chooses the tiger. Does this mean the user knows what a tiger is? Well only to a degree, it means the user knows a tiger is an animal with four legs. that's about the most we can deduce from that test. Then we show pictures of a dog, an elephant, a cat and a tiger. User again chooses the tiger. See this test was much better (harder), the user had to know the specific features of tigers to get it right.

Relevance to Pleco. Some characters are similar to other characters, some are completely different. The harder test you can present someone will give you a set of characters which are similar to the test character. I get a lot of tests right but I still haven't properly learnt the character because I'm usually presented with options which are very different to the test character, basically it's too easy.

Imagine having to pick 岂 from the following...

岜
岂 (hard)
击
屯

versus

你
岂 (easy)
删
日

Well if you know the composition of all characters then you can (relatively) easily pick out characters which are similar to each other.

Say you have a Character Similarity Score (similarity to an object char), it's a function of Stroke Count, Primary Radical, Sub-Radicals. Given a test character you derive a list of like characters, pick some at random.

Some quick points re this

- With radicals you don't always have to match exact, some radicals are very close in appearance and could consider so in this calculation, the more like the greater it adds to the score. Kind of same with stroke count, if char has 7 strokes then consider 6 and 8, but they add less to the similarity score.
- Perhaps better to increase this specialisation as character is learnt. If I don't know the character then I don't need it, in fact it might even be detrimental at first, best not to convey information via the test structure (or hint to put it another way).
- Going even further you could even consider the style of the character, curved lines vs straight, intricacy, stuff like that, but of course you need the reference info for anything you are going to incorporate. Given there aren't too many radicals this might not be so hard to categorize.
- For multiple character tests consider jumbling characters, replacement etc (with original and similar). Derive options from the test word, make the user hone in on the detail. I will often get a test right because I know one character in the word, 你+char is too easy when the other options don't have 你.

Which one is nihao?

好尔
你好
你女
尔好

- Following from above (this might be going a bit far). The 4 seasons end in Tian so I'm always getting them right because of that even though I'm not so sure about the first character. Well if the test could recognize that I have already learnt Tian, then it could make that constant and only change the first character. Or it could look into word list and find words that also have Tian as the second character. Make constant that which is already known.
- Actually it would seem you could already do something pretty similar to this given that the Handwriting recognition provides a variety of similar characters.

BTW I'm a programmer so I know it's not always as easy as it sounds. People are already complaining about how complicated it is (actually I made a remark).

Cheers.

 

[mikelove]

Hmm... well there are a couple of possible approaches to this problem.

The simplest (and perhaps most Pleco-2.0-like) solution to this would be to create a new flashcard construct which allowed one to create groups of similar / easily-confused cards to be frequently tested together - something sort of like categories, but smaller and more simply organized, which would be specifically tested against each other in multi-choice questions.

A more complicated solution would be to automatically match cards based on basic information we have about them - this wouldn't really work well for similar-looking characters, but it could certainly work for Pinyin variations, homophones, and words with one of their two+ characters in common.

The really advanced thing would be a system of character similarity scores like you describe. We have some data that we might be able to mine for this - radical, stroke count, and even now some other non-radical components / positions of the radical and of those components, but to really blow people away with this we'd probably want to go through and specially generate our own set of character similarity scores. This would actually be a great little project for Amazon Mechanical Turk - show people a character along with, say, 10 other characters that might be similar and ask them to pick the most similar-looking one (or rank a couple of them); run through a million or so of those and we'd have ourselves a pretty good data set.

 

[Mark Kang]

四十多页，我都看了一遍。Mike,你真是皇帝！

A special linking function for easily confused characters I think would be useful to a lot of people. In my mind it needs to be user defined to really solve the problem for me though, but it would be a job I'm sure. So, for my birthday then, how about a simple one? I desensitized my screen on my old Tungsten E from doing flashcard tests. Win Mobile allows for switching to landscape and Pleco also has a left hand option, BUT it seems that my OK button up in the corner there really is getting a lot of use, not to mention it's kinda cramped up there in either view. What do you think about putting one on the Pleco screen where it's appropriate？Make it round, or even in the shape of the Pleco fish if you want .

And, I 很积极的second the request for a card search toggled off the main dictionary search field. The current card search is cumbersome for simple searches based on character and/or pinyin.

加油！