Flashcard Managment in Pleco 2. (Question)

[Smoodo]

I've been using the 'Koreth' method of repetition spacing referenced in a thread below. I'm very happy with it and have carved out a sizable chunk of the HSK list, but now I have a management problem.

posting.php?mode=quote&p=3613

Scenario

Imagine this, I've been studying Chinese for a couple of years now and my proficiency in listening, speaking, reading and writing are all at different stages; from best to worst in that order. Now, I'm at a university in China going through 6 different courses that are running me through easy parts and tough parts to even out my Chinese. The course materials that I use also have words that I must learn to keep pace with the class.

Use case

I've been studying for the HSK and using one of the four lists from this site and have gone through the first 1000 cards and they have a pretty long history. Duplicates are coming up in my other classes because they are a similar level or previously learned materials. I add them to a separate list because I'll have a listening and writing test over them. They must be in my head that day, so it makes sense to keep them separate for special review.

Desired Solution

I need a way to merge the list of my class characters into the HSK list and make the statistics additive and take the latest date reviewed. It would be fine to take the flashcard 'rank' that is highest.

Fundamentally, the problem is that the same card has two histories and one may reflect the truth more than the other, but both together reflect the true exposure and practice with a 'word'.

Database Question

Does the future SQL lite database that's coming have a key in one of the rows that will identify the dictionary that should be used for the definition?


I realize the squeeze you deal with when having to not let intellectual property from your dictionaries leak out through flashcards, but some definitions are 'better' than others. It would be nice if there were some portable (flashcards are shared) way of having a flashcard list that would remember the dictionary preference.

In the case that someone doesn't buy all the dictionaries, Pleco could default to whatever available dictionary has a definition, but otherwise restore the links.

Thoughts, obstacles to upgrade, solutions

I've spent so much time reviewing over a thousand words potentially hundreds of times that the cost of loosing that history is very expensive. If for some reason Pleco 2.0 could not use and understand the history from Pleco 1, I could not upgrade without some kind of solution. History carves out the known from unknown in these comprehensive lists floating around out there. I would certainly contemplate a "knowledge" history in Pleco that spans all potential lists (with a flag per list). If it's too much trouble an alternative solution would be for Pleco to provide a delete key in some sort of cleaning review mode for those massive lists that would allow me to delete entries that I feel I've mastered. This way I could use comprehensive collections and not start with absolute zero with basic characters like 'ren' (person).

I see creating special lists to keep things in mind for a test or what not, but also tracking a total knowledge base. The knowledge base may get really large and repetition spacing would put character's review dates farther and farther out. I imagine a case where Pleco thinks because you've gotten it right 400 times that you have that baby mastered, but in fact it's been so long ago since you saw it that maybe you did forget it. I think our minds can effectively forget anything given enough time, no matter how much review.

My suggestions revolve more around database design than rough programming tasks. If the database and keys are setup properly these should be easy to implement. Very few simple rules, and good descriptive tables columns should be adequate.

Thanks for reading and considering this suggestion.

 

[mikelove]

On the first suggestion, our basic philosophy on this is that since cards can now belong to multiple "categories" instead of being tied to a specific "list," there's not really any reason why you'd want to have two copies of the same card. Even for creating a special class review list like you describe, you could simply create another category just for that, add all of the cards you wanted to review to that category, and select only that category for review. So there would only be one set of statistics for a particular card, and hence no need to merge them.

We do actually support multiple "scorefiles", so a card could still have multiple statistics associated with it, but those are mainly intended to allow you to track different types of review differently, stroke order versus Pinyin versus tones specifically and so forth, so we're not planning to allow you to merge scores from different files together.

Re your database question, our new XML-based flashcard export/import format does indeed include dictionary associations; our goal is to be able to have someone dump their flashcards to an export file and have them reproduced with something pretty close to 100% accuracy when imported, so virtually every significant bit of data about a flashcard is recorded in the export files. (though there are also options to disable portions of that, so for example if you want to give someone your word lists but not also have them inherit your card statistics/history/etc, you can easily do that)

 

[Smoodo]

What about Pleco 1 history?

How will the Pleco 2 handle my Pleco 1 lists and histories, including duplicates? Will it eliminate dupes and collapse all cards into categories of their list name, keeping the review history?

 

[mikelove]

I actually hadn't given much thought to that... the flashcard converter program won't be out until the finished 2.0 is ready (since the flashcard data file format may change during the betas and we're planning to actively discourage people from relying heavily on the beta-version flashcard system), so it's not even really on our radar yet design-wise. It may eliminate duplicates, in which case we'd probably convert the various lists it was duplicated in into category assignments, merge the mergeable portions of the statistics (last reviewed and total times correct/incorrect), convert the highest rank into an equivalent "score" and discard the un-mergeable stuff (like the detailed review history). Or it may leave duplicates intact and allow users to sort out which version they wanted to keep on their own.

 

[Smoodo]

Curse of flashcard value.

The hard core posters on this site seem to have huge lists, with histories, that they've been building over a long time.

When I first became a customer I purchased a dictionary for my PDA. When I returned as your customer, I bought a flashcard program with dictionaries. . I can buy great dictionary PDA's out there, especially in China, but they don't have a flashcard part, where I see the real future value (to me).

Learning Chinese is an ongoing effort for me. Resetting to zero is out of the question for me. I've fed the flashcard system well and now I'm reaping the benefit of smart review sessions at exactly my Chinese level. That is too valuable to discard because it took hundreds of hours to build it into PlecoDict.

If you take that into consideration, what reason would people like myself ever have for upgrading to another version? I totally dig new features, cool user interfaces and more dictionaries, but at the end of my day there is the greater purpose of learning Chinese and not just using PlecoDict as a reference tool.

This is the curse of flashcard value. Without flashcard and history, you are in a race to be the best collection of dictionaries out there. These dictionaries are controlled by people/corporations that you cannot control. Flashcards are unique Pleco intellectual property of which you have complete control regardless of what dictionaries are out there.

 

[mikelove]

Sorry, I should have been clearer. I wasn't talking about resetting everything to zero, I was simply talking about dropping the data that can't be merged for cards which are being merged. The detailed card history is just a string of 1's and 0's, like this:

10000101111101111101

Last answer correct, four answers before that incorrect, answer before that correct, answer before that incorrect, and so on. Since there aren't dates attached to any of those answers, there's no possible way to merge two of these histories together. Counts of correct/incorrect answers are easy to merge, you just add the numbers together, so we'd certainly be keeping those along with most of the other data, but since it's mathematically impossible to merge answer-by-answer histories together, I don't really know what you expect us to do with those. More practically speaking, they're only used with the "in a row" and "of last" rank promotion calculations, and with our new scoring system in 2.0 those calculations are no longer going to matter to most people (since the default behavior will now have the score changing a little after every single answer), so it's not really a big loss.

As for the flashcard converter program not being out until 2.0 is finished, this is simply to prevent people from rushing out and converting their entire flashcard collection to the 2.0 format the second the first beta comes out, deleting all of their old-version backups, and encountering a bug in 2.0 that causes their flashcard collection to be corrupted / deleted. It's also to save us from having to make sure that flashcard files created in one beta version will work correctly in the next one; it's very likely that at some point along in the beta process we'll need to change something about the flashcard file format, be it for performance optimization, bug-fixing or a last-minute feature addition.

 

[Smoodo]

Opps.

I said history, but think that what I mean is the total statistical meta-data for each card.

It's valuable to know when the last time I reviewed it was and interesting to know how many times I've gotten it right or wrong. It's rank determines it's frequency of testing occurrence.

 

[mikelove]

Well that stuff should all be preserved, since it's so easy to merge. So no worries there.