It's probably an issue with the text encoding you're saving in. I had a similar problem a while ago, which I solved by switching on an option that allowed me to choose an encoding every time I saved a text file.

Yea I reckon you should find a way to save in UTF-8 encoding and that'll probably solve your problem.

Yeah, it was the UTF-8 thing. Excel didn't have that option (only UTF-16), so I used NeoOffice.

Thanks for the quick reply!

You might be able to just load the file in Notepad, and choose to save it as UTF-8.

Hello Resolve,

I just downloaded your program. Looks sweet, course I'll have to start all over, BUT some folk said if I import, my list order won't be changed. I exported from my flashcard program, opened with Word, exported to txt UNI 8 and imported into program and then synched. I logged on from pc and cell and for some odd reason it starts me off on card 1604. Any clues?

TIA!

wow, I can't believe I discovered this thread so late! This seems to be the program I was looking for. I'm on a mac and I've explored many options for flashcard programs.

I've used iFlash, which is brilliantly designed, but I just can't get myself to stay with it. I installed Mnemosyne on OSX (I wrote the guide on how to do it) but mnemosyne's simplicity is frustrating sometimes. I also bought a license for Supermemo and ran it in Parallels, but that's just more of a hassle than it's worth and Supermemo is not the most user-friendly program around.

Consequently, I'm really excited about Anki, especially the fact that it was designed with the Japanese-learner in mind (Full UTF-8 support!). Thanks for the great piece of software; I'm sure that it will grow to become an amazing application.

Furthermore, I would like to second the sentiment that Anki shouldn't be focused on RTK as this site is perfect for that. Instead, I think vocabulary/grammar and sentences would be a more apt application.

If you need any testers on OSX I can gladly offer you my services, as well as compile for you if necessary.

I agree!

I used excel@japanese strictly for the jlpt last fall and it helped enormously. A friend said I studied kanji like the Heisig system and I checked it out in January and got hooked.

Read about flashcard software and found a free, nifty little program called Memory Lifter and have used that til now, at around 1200 now. Problem is last year I bought a 17" widescreen notebook and lugging that thing around is a bear. Thought about the PDA route and then made little html files for my cell phone, but the files were always too big. So, I've been hanging tight til I could update my cell to a document viewer. And now there is Anki.

My problem now is that I imported my files from ML into the software and then synched. On the PC it's in order, but internet and cell access starts at 1604 for some reason and I'm not there yet. I wonder, should I have imported directly to the site and not the software? I don't think it should matter, but it might. Before I erase them all, anyone have any thoughts how to solve this?

Sorry, can't answer Aikiboy's question.

I've been thinking...as opposed to studying. And my thoughts so far today include deciding not to start using Anki (yet). It looks great but I reckon Fabrice has something in the pipeline that will be worth waiting for...so that's what I'm going to do...wait.

Aikiboy Wrote:My problem now is that I imported my files from ML into the software and then synched. On the PC it's in order, but internet and cell access starts at 1604 for some reason and I'm not there yet. I wonder, should I have imported directly to the site and not the software? I don't think it should matter, but it might. Before I erase them all, anyone have any thoughts how to solve this?

The designer recently modified the standalone program to present the cards in the order they were added, but I guess the web/cell version just gives them in random order like always. He's working on a much snazzier version of the web review system that'll probably incorporate that modification, but dunno precisely when that will happen.

For now, you could delete all the cards that you've added and split the import file up into smaller pieces. Maybe you could divide it up by lessons. Then just import the smaller files as they become necessary.

I'm still a little confused why people are using Anki to study RTK. This site here does the job pretty well, and Ricardo made a little script that makes this site accessible from a lot of portable devices.

snozle: Thanks for the offer. Hope you find find it useful! I've come across your OSX installation instructions before when looking into other flashcard applications. It looks like a bit of a pain to get running

Synewave: That sounds like an excuse for skipping study! ;-)

Aikiboy: The problem is that the syncing algorithm doesn't preserve the order of the deck. This is hard to do when you consider that cards can possibly be added on both sides before syncing again. Still, I'll have a look and see what I can do. In the mean time, you can log in to the web interface, then go to http://anki.repose.cx/deck/add and upload the deck from your computer. This should preserve the order and you can then use syncing to keep the two in sync.

Note that adding new cards after that and syncing will result in the new cards being in a different position on the server. Will that be a problem? I agree that the kanji should be learned in order, but I don't think they should be practiced in order, or it becomes too easy to answer questions (as you know that the next kanji will probably be based on the last one.).

good program, I have now imported all of my flashcards from mnemosyne and I am going through re-learning them again. it's actually been quite good to see how much i actually do know. For example I get to certain cards that I just know instantly because they have been drilled into my head for so long.

One more suggestion:

Again, with mnemosyne (current program of choice) there is a "categories" section that means you can enable/disable certain parts of the collection for specific studies. something that i would love to see implemented in Anki.

EDIT: typo

Resolve, excellent program. Very clean and easy to use. On another note, does anyone have any 1kyuu or higher decks? I would love to get them if possible.

One of my favorite things about the progam is that the reviews are scattered throughout the day, and I can see the cards coming. It's easy to feel overwhelmed when a whole bunch of cards expire at once, but with Anki, I can review 160 cards every day and feel like I'm hardly doing anything.

Yorkii: the next major release (0.3.x - probably 2-3 months away) will feature a completely new way of handling cards which will allow for tags/categories as well.

Hyland: I have full decks of JLPT1-4 that were generated from edict. The problem is that the definitions can be very long, or not specific enough (it's very difficult to study English -> Japanese when you didn't make the definitions yourself). There are also a few mistakes where the automatic matching of kanji to a definition chose the wrong definition. I strongly recommend using Anki to input words you encounter yourself (from a textbook, from websites, etc). Doing so means that you a) know how to use the word in context b) can remember more easily because the word holds some relevance and c) you can make useful, accurate definitions instead of using a catch-all dictionary entry. These days, instead of studying words by themselves, I often add whole sentences and study them instead.

The sample decks provided with Anki are a subset of these full JLPT decks, with the edict definitions replaced with the definitions from my personal deck - so the definitions are generally clearer, but some of them may seem strange as they only hold relevance to me personally.

So basically I don't recommend using a pre-made deck. If you want to know how to actually use the word you need to look up examples anyway - and in that case, you might as well just study via a JLPT word list, looking up the words and typing them in yourself. But I realise that many people don't want to go to the trouble of entering in cards themselves, so I provide the sample JLPT2-4 decks with Anki, in the hope people will see how Anki helps improve their memory, and then start adding their own cards. I don't provide the full JLPT edict-based decks because the quality of the entries makes it pretty difficult to study. But, if you want to give it a go anyway, you can get them from http://repose.cx/anki/jlpt.fc/

After two days of using Anki (I know, not that long) I have several questions/suggestions:

1. Is there a way to choose when/where to display the hiragana? I inputted some sentences with Kanji that I don't know anticipating using the hiragana to read until I have completed RTK1. However, when Anki displays it Kanji --> hiragana + english. This is fine, and even preferable, for when I've finished RTK1, but for now I can't read all the kanji.

Should I just go hiragana to english for now and add Kanji when I complete RTK1 or is there a way to change this?

2. I read the explanation of the Leitner approach on the site, but I was wondering if it's a strict Leitner algorithm or a more advanced one akin to the ones found in Mnemosyne and Supermemo? The number grading system seems to indicate the latter, but I haven't had enough experience with the program to see if they always move the card to a prescribed date in the future or if it learns from your memory.

3. In addition the the categories, which you said would be in the next major release, I can suggest checkboxes next to each card to enable/disable it without necessarily deleting it, possibly to save it for future use.

I recommend you try just using kanji -> hiragana and english. I think you'll be surprised how quickly you remember kanji even learnt outside of RTK. You can use those sentences for just reading practice and actually study writing via RTK, so it's not really doubling up.

I don't think creating hiragana on the question side is a terribly good idea because it'll mean a lot of work going through cards and editing them in the future when you do actually want to practice the kanji. But if you want to do that, you can do it by creating a text file in the format:

kanji (kana) : english

and using Anki's import menu item.

2: the algorithm is most similar to mnemosyne's. Anki used to use an implementation of the supermemo v5 algorithm, but I ditched it because it had a tendancy to 'overlearn', leading to suboptimal behaviour (cards jumping from 30 days to 300, cards getting stuck on 1 day repetitions for multiple repetitions). Supermemo's latest algorithm is more refined, but it still exhibits the same problems - if you're not totally consistent and you add stuff that is a bit easier or harder than normal, it'll start to get confused.

So Anki uses a simpler algorithm, that is a lot more flexible and precise than Leitner, but not as complicated (and thus less prone to error) as Supermemo. The algorithm is still developing and more 'learning' may be included in the future, but I think that too much learning is a bad thing, as it leads to the program trying to interpret patterns in your study which don't need interpretation.

Interestingly enough, the Supermemo authors wrote a prototype Supermemo that used neural networks to do the scheduling, but they dropped it because they said it was too prone to confusion. I think Supermemo itself suffers from the same problems, though the parameters are better controlled.

The ability to enable/disable repetitions for certain cards would be nice, and is something that will be implemented at one point.

yorkii: it looks like there's a problem with syncing for you. Could you email me your deck so I can investigate? Also please let me know if Anki throws up any errors when you exit (on win32) or in /dev/console (on mac)

yea, i was messing around with trying to work 2 decks at once last night and it made a whole mess of my collection and what i had and hadn't learned. ill write more details in an email to you.

resolve Wrote:the algorithm is most similar to mnemosyne's. Anki used to use an implementation of the supermemo v5 algorithm, but I ditched it because it had a tendancy to 'overlearn', leading to suboptimal behaviour (cards jumping from 30 days to 300, cards getting stuck on 1 day repetitions for multiple repetitions). Supermemo's latest algorithm is more refined, but it still exhibits the same problems - if you're not totally consistent and you add stuff that is a bit easier or harder than normal, it'll start to get confused.

So Anki uses a simpler algorithm, that is a lot more flexible and precise than Leitner, but not as complicated (and thus less prone to error) as Supermemo. The algorithm is still developing and more 'learning' may be included in the future, but I think that too much learning is a bad thing, as it leads to the program trying to interpret patterns in your study which don't need interpretation.

Interestingly enough, the Supermemo authors wrote a prototype Supermemo that used neural networks to do the scheduling, but they dropped it because they said it was too prone to confusion. I think Supermemo itself suffers from the same problems, though the parameters are better controlled.

I'm using Supermemo currently, but I don't like the way it overlearns so easily. It very quickly and recklessly jumps off the normal curves if you start a collection with especially easy items. Then, because of the way it picks the schedule ahead of time, even after it has got more sensible curves you still have to wait until you finish all the 'badly scheduled' reviews before they get a good schedule.

Strangely, if you set a forgetting index of 10% and you are very good to remember your items, SM will actually try to make you forget 10% of your items by scheduling them very far away . (Well.. technically its goal is to produce the minimum number of reviews to get meet the FI, which has the effect of making you forget about that much).

It would've been better if instead of having a target forgetting index, you just tell it how many reviews you want to do per day and it does its best to minimize your forgetting.

I'm interested in giving Anki a try, but my issue is that I've been using Mnemosyne for about 10 months and have almost 5,000 cards in it. They're showing up at a manageable rate of about 80 a day at the moment. If I just import them all into Anki, and start from scratch, it will probably be months before the cards settle into the right review frequency based on how well I know them.

Is there any way to import into Anki where you can give it an idea of how well you know it? Ideally this would be based on the xml format that Mnemosyne produces, but I might even be willing to write a perl script to convert the Mnemosyne xml file into an Anki readable format if that was possible.

thanks!

I like the idea of the timer function in Anki; however, in practice it doesn't work that well for me. I review at different times each day and Anki sometimes won't show me cards because it is waiting 24 hours.

For instance, last night I reviewed my cards and it said that the next card will be displayed in 24 hours. Today I would like to review this morning instead of at night, but the cards won't be available until tonight. Is there any way to have it revert back to expiring cards at midnight?

ergerg: implementing an importer for mnemosyne should be pretty easy. could you post or email me your deck, or a subsection of it, so i can see the output format?

snozle: I think in the long term having to wait a few more hours for the cards to appear isn't a big deal. Personally I find myself only using 'review tomorrow' for the really hard cards. Anki has no 'reset at midnight' feature currently, but you could change the interval for '1' to 18 or 12 hours with about two lines of python code if you wanted to. Perhaps in the future I'll add a 'step into the future' option which allows you to pull cards in from the next 12 hours.

resolve Wrote:ergerg: implementing an importer for mnemosyne should be pretty easy. could you post or email me your deck, or a subsection of it, so i can see the output format?

If this can be done, be sure to let me know, as this is the exact same thing I am going through now. I imported my collection - a mere 1300 or so words into Anki and am going through them like 300 a day to try and get the program to sort out which ones I know and don't. If I could just through my mnemosyne file in and all the learning data me kept, I would be on cloud nine

The xml file looks fairly straightforward (and the source code for Mnemosyne is available) so I don't think this will be too hard, but I don't have time at the moment to look into it any real depth.

The xml files contain lines like:

<item id="e034b595" gr="2" e="2.333" ac_rp="4" rt_rp="9" lps="3" ac_rp_l="1" rt_rp_l="0" l_rp="71" n_rp="73">

Looking at the code:
gr -> grade
e -> easiness
ac_rp -> "acquisition repetitions"
rt_rp -> "retention reps"
lps -> lapses
ac_rp_l -> "acquisition reps since lapse"
rt_rp_l -> "retention reps since lapse"
l_rp -> "last rep"
n_rp -> "next rep"

I don't think it's important (or probably even possible, given that the algorithms are different) to reproduce the Mnemosyne schedule exactly, but just using the repetitions since the last lapse one could probably do a decent half-assed job of putting the items into some reasonable kind of schedule.

Yorkii: how are you liking Anki? Is it that enough better than Mnemosyne to be worth making the switch?

thanks!

it has it's pro's and con's at the moment.

I know that the developer, Resolve on this forum, is working on Anki and it is still in its very early stages so of course I can see the potential in it.

Things that I like better in Mnemo are:

1. Add Cards: extremely simple -- drag and drop copied text from the net and the formating stays the same (line breaks issue mostly.)
2. Edit cards: can be done on the fly, again, with a really easy interface.

Things that I like better in Anki:

1. I have a concrete image of my schedule. e.g. graphs etc.
2. My learning data is stored so I can see how well I am doing etc.
3. (this is a big one for me [if i can get it to work!!]) The syncing between multiple machines and Japanese mobile phone based support is awesome!

ergerg> keep me informed if you find away to get the mnemosyne file imported with some of the learning data retained... I know nothing about programming and stuff so any help would be great.