|
 |
Flash Card Software | ||
Tags: Flash cards | Software | ||
Share with:
| ||
 Language Learning Forum : Language Programs, Books & Tapes |  | |
| laxxy Triglot Senior Member United States Joined 7121 days ago 172 posts - 177 votes  Speaks: Ukrainian, Russian*, English Studies: Japanese | Message 57 of 67 14 January 2007 at 12:45pm | IP Logged | ||
|
that's a pretty fast growth, I wonder what the first three intervals look like.
Some of these programs (Mnemosyne and Twinkle definitely, maybe some others?) are open source, so you can tweak the algorithms if you feel a need to. 1 person has voted this message useful | |||
| virgule Senior Member Antarctica Joined 6842 days ago 242 posts - 261 votes Studies: Korean | Message 58 of 67 14 January 2007 at 2:31pm | IP Logged | ||
|
@Thomasus: I don't want to argue about different algorithms or approaches. By transparent I mean that the last time I checked, the FullRecall documentation didn't state how it calculates the next interval of an item. It's not about how complicated the maths is, it's about whether the user can know. That's how I can decide whether I find the assumptions behind the program reasonable.
I don't think a good testing of algorithms would be that hard. The problem is that I can't get motivated to do it, because given the vendor lock-in we experience, I am not going to change anyway. With testing the different applications on a single person, external effects should not affect any one of the programs differently. Because I would allocate the items randomly to different algorithms, their average will be the same. For the same reason I suggested a relatively large number of items (2000). Of course, ideally you would repeat the experiment with different people (each trying the same few applications) to get more reliable results. The point of motivation is really that I don't have a list of 2000 or so items that I wish to memorize. I do wish to know, but not to the extent that I will memorize 2000 words from a new language, or random associations. I would be interested in what you found when trying to compare the efficiency of different flashcard software. How exactly did you set it up; how long did you run it for; and what did you find? @Gidler: A unified data format would be difficult to implement, simply because different programs use such different approaches. How would you convert data from a neural network into a Leitner system? What do you do with the extra item-level information when converting to a pure Leitner system? Where do you get item-level data from when converting from a system where there is none? How do you deal with sounds and images? What might be more realistic is converting between similar programs. The use of XML in some OpenSource formats might help a great deal with that. 1 person has voted this message useful | |||
| tomasus Pentaglot Groupie Czech Republic Joined 7034 days ago 54 posts - 56 votes Speaks: Slovak, Czech*, EnglishC1, German, Russian | Message 59 of 67 14 January 2007 at 2:52pm | IP Logged | ||
I didn't. I have used Supermemo on and off for maybe 8 +- years, and never felt compelled to compare efficiency, as as far as I can remember, there was never anything comparable. The only reasons I tested other flashcard programs was for technical reasons (different functions, better unicode, better GUI, and so on). The only attempt at testing was made against Fullrecall that appeared only relatively recently on the scene, but after I realized how hard it would be and how the results still might be useless, I have given up. (For one example among many others, Fullrecall targets 100% retention, in Supermemo you can setup forgetting index from 3 to 20%, these two are hard to setup equal given the scales and the real meaning behind them). Edited by tomasus on 14 January 2007 at 2:53pm 1 person has voted this message useful | |||
| gidler Senior Member Finland Joined 6625 days ago 109 posts - 118 votes 1 sounds Speaks: Finnish* | Message 60 of 67 15 January 2007 at 12:50am | IP Logged | ||
As I said, the format should contain data of a lower level. Timestamped events like "added an item", "showed an item to user", "user graded an item" and nothing more specific than that. A flashcard app could then process the log and simulate the events as if the user had done all the repetitions with the importing application. The intervals wouldn't of course be the same, but good flashcard apps can handle delayed or early repetitions and still do their job. As for images and sounds, they're not that big of a problem as long as you don't set too many requirements - but indeed it would be sensible to first make it text-only and expand later. 1 person has voted this message useful | |||
| hagen Triglot Senior Member Germany Joined 6962 days ago 171 posts - 179 votes 6 sounds Speaks: German*, English, Mandarin Studies: Korean | Message 61 of 67 16 January 2007 at 1:40am | IP Logged | ||
|
Designing a unified data format for flashcard files shouldn't be a problem, e.g. an XML format where every software could store their special data along with the basic data (though you would have to think about how general you want to be, e.g. in allowing n-sided cards).
I can't imagine that it would easily gain acceptance among the bunch of flashcard programs though. And any decent software should already have a plain export mode anyway (which is how I recently got my cards out of VTrain and into my own program). 1 person has voted this message useful | |||
| luke Diglot Senior Member United States Joined 7207 days ago 3133 posts - 4351 votes Speaks: English*, Spanish Studies: Esperanto, French | Message 62 of 67 21 January 2007 at 8:10pm | IP Logged | ||
|
I've employed a couple of techniques lately with mnemosyne for words that are having difficulty sticking to my mind. They are more for managing the learning process than for learning per se.
1) If a question/answer is not getting easier with several repetitions and you think it's important, press <3> for short term mercy. That takes it out of the lineup for the current session, but it will be scheduled in the near future. <3> is for "I remembered it with difficulty - the interval was too long". 2) If a question/answer is not getting easier with several repetitions and you think it's something you can put off until later hit <5> for "I know this really well - the interval was too short". Both of these techniques are for the student who creates a gazillion cards, but cannot remember them all yet. With these techniques, you can still create new cards and not be overwhelmed in the sea of the unknown. Hopefully the newer cards you create will be more memorable, or at least things you want to learn in shorter order than those you postpone. Edited by luke on 21 January 2007 at 8:12pm 1 person has voted this message useful | |||
| tomasus Pentaglot Groupie Czech Republic Joined 7034 days ago 54 posts - 56 votes Speaks: Slovak, Czech*, EnglishC1, German, Russian | Message 63 of 67 22 January 2007 at 4:02am | IP Logged | ||
|
Doing the above in FR or SM would spoil the learning process. Doing it in Mnemosyne might probably work (ie. not cause much harm), however Mnemosyne has an option to upload detailed statistics about your learning process to their server, for research purposes. I would recommend to turn that option off, especially if you use the above techniques on a significat portion of your database items. So that their research is not spoiled by this data (albeit in the total sum it would probably be an insignificant portion). Edited by tomasus on 22 January 2007 at 4:03am 1 person has voted this message useful | |||
| gidler Senior Member Finland Joined 6625 days ago 109 posts - 118 votes 1 sounds Speaks: Finnish* | Message 64 of 67 22 January 2007 at 6:10am | IP Logged | ||
|
As tomasus said, please make sure you have the statistics uploading disabled. Also, Mnemosyne *does* keep item-based difficulty scores, so by doing what you described you're messing up the scores. (But the algorithm uses the scores only for adjusting repetition delays so anything catastrophic isn't going to happen.)
I can't stop myself from mentioning that in SuperMemo, you could manually adjust per-item "forgetting indexes" to increase/decrease repetition frequencies based on importance. The default forgetting index can also be adjusted to change average retention. This allows you to increase the speed of learning new items without getting an overwhelming number of pending repetitions (at the cost of decreasing the amount of information that actually stays in your memory). More details here. 1 person has voted this message useful |
|  | 
Printable version
|
| Copyright 2024 FX Micheloud - All rights reserved No part of this website may be copied by any means without my written authorization. |
