I'd like to write about my experiences with review software. I've tried most of them out there, but I mainly settled on Mnemosyne and later SuperMemo.

For anyone interested, I've made RTK1&3 lists that can be imported in Mnemosyne or SuperMemo: you can msg. me if you want copies.

I'm a software developer so I studied Mnemosyne's internals.

I like Mnemosyne's method of reviews scheduling. I would say was a self adjusting Letiner system. Its actually based on an old version of SuperMemo's algorithm, but somewhat tamed down. It adjusts interval ratios only slightly at each review (SM makes bigger adjustments, for better or worse). Its enough of an adjustment to make it better than simple Leitner IMO. Mnemosyne also throws in a little randomization to the review intervals. This has a good effect on spreading out your reviews so you don't get a huge pile to do on one day.

What I don't like about Mnemosyne is there is very poor visibility of progress, and statistics and stuff which I really like to see. It's also cumbersome to add formatted Q & A, and find stuff that you've added to its database.

SuperMemo is my current tool of choice, although its far from perfect IMO. SM adds some levels of complexity that allow it to adjust to each user's personal ability & each database of questions. This can really improve the efficiency of learning in certain cases. However in my experience this is both a blessing and a curse because it over-learns and over-adjusts. When I already know some of the information quite well, for example I already know about 600 RTK1 kanji, I can remember them really well. In this instance SuperMemo thinks I'm some kind of genius who can remember new information for weeks without a single review. This makes reviewing the other 1400 RTK1 new kanji really hard.

For example it currently thinks that I can memorise 3600 items/year/day/min, so for every minute reviewing each day I can memorise 3600 items a year. If only that were true I could be a real Japanese guru in a year. SM adjusts itself over time to correct this, but in between there can be a few weeks of frustration if you don't know how to deal with it.

Mnemosyne's developer has stated that he doesn't think the added complexity of SuperMemo's algorithm is good. At this point I tend to agree; for a collection of questions of which you already have some familiarity, SM over-learns.

Despite this (IMO) flaw in its algorithm, I really like all the statistics in SM, the visibility of what I've done and what I'm going to do in the future, the ability to organise information, the editing abilities for Q&A; especially for adding new vocabulary and phrases its easy to cut & paste from the internet.

I've learned that in order to stop SM from *going nuts*, I sometimes pretend that I don't know some easy questions. This makes it think I'm still human; not some alien from Mars with 2 brains. What would be cool is if it was possible to set up (human) boundaries so that it didn't over-estimate one's abilities so easily.

If I was SuperMemo's developer... I would add an optional setting that said how familiar newly reviewed items are to me on a scale of 1 to 5. I could set ichi, ni, san... to 5, somewhat familiar items to 3 and brand new material to 1, and the algorithm can adjust appropriatly for each level. If I have some time... maybe I will write such a tool .

very interesting write-up vosimiura, thanks! I purchased a license to supermemo after reading rave reviews and I've never really used it. I only have macs at home and though I run windows in parallels and bootcamp I find it tedious to switch simply to review flashcards.

I currently use Mnemosyne for everything but the Kanji, which I use this site for, and find it suitable. However, in the back of my mind, I always wish I would take the time to learn supermemo because I keep thinking that it has so much to offer.

If I read correctly, I think you would probably recommend I stay with mnemosyne and I think you have convinced me to do so. Supermemo, though intriguing, is just not worth the extra effort.

Prove me wrong...

i tried supermemo,

but with all the lack of unicode support and hence "Not for Japanese studies" I gave up pretty quickly. also, if you make cards in supermemo they are huge. I made a file of about 1000 flashcards as a test and the folder for that collection was about 80MB if i remember correctly.

I really can't see there being a HUGE difference between using mnemosyne or supermemo. it's a case of　隣の芝は青く見える！

you have to look at it this way: 30 years ago there wasn't either...!

I think if you're happy with it you should stick with it.

Like I said, I wanted more statistics and organisation and stuff. I'm just that kind of person; I look at SuperMemo statistics and graphs quite frequently and that makes me satisfied.

I will give SuperMemo a fair shake and use it for a long time to see if its algorithm is really improved over other tools. I'm not yet convinced, but I will listen to the hype and give it a chance over a long term.

How's Mnemosyne working out for your memory?

without mnemosyne i would be less than half as proficient in Japanese as I am today.

yorkii Wrote:i tried supermemo,

but with all the lack of unicode support and hence "Not for Japanese studies" I gave up pretty quickly. also, if you make cards in supermemo they are huge. I made a file of about 1000 flashcards as a test and the folder for that collection was about 80MB if i remember correctly.

They don't take very much space for me. My RTK collection with all 3006 entries from RTK1&3 takes 1.2MB, and my vocab collection with 2000 items takes 2.5MB.

The Q&A are just HTML so you can have Japanese in them no problem. I got SM2006, which has more support for Unicode than 2004, for example the Contents view can display Japanese for item names and categories. I understand from other posts that you review japanese -> japanese so that is an essential feature.

A feature I like is that to add/edit questions you just click on them and you edit them WYSIWYG, plus I just paste stuff direct from internet explorer and it gets the formatting too. So for example if you paste an entry from your favorite online dictionary it will keep the formatting, colors, etc just like the original.

Its also easy to create cloze type questions, like fill in the gaps. You paste in the sentence, select the part you want to omit and very easily SM makes a question and adds [...] (in yellow) where the missing word goes.

Mnemosyne seems to be working fine for me now though I haven't really been using it that long. I just got it running on my mac, which took an entire day of trial and error. Incidentally, I wrote a guide <--shameless plug, detailing the process.

I love the concept of an SRS in general, the fact that I keep acquiring material and don't have to worry about reviewing it, just let the software do it's thing.

My question for you, who seem more experienced than me when it comes to SRS usage for Japanese, is what exactly do you do for vocabulary? Do you use individual words and translate them to English or do you do the "sentence method" like outlined at All Japanese All the Time? I kind of agree with khatzumoto, the author of All Japanese All the Time, when it comes to vocabulary in context; however, the problem that I have with his method is that he says not to start until after you have finished RTK1 because he encourages the sentences be from kanji to kana (japanese to japanese). That's great, but I feel that I should be doing something while I work through RTK.

Any suggestions?

snozle Wrote:I love the concept of an SRS in general, the fact that I keep acquiring material and don't have to worry about reviewing it, just let the software do it's thing.

Sorry, I looked over this thread, but I couldn't imply what SRS is. The concept of "not having to worry about reviewing, just let the software do it's thing" is certainly quite interesting.

EDIT: Oh, yeah...it hit me. Like KBAM, how could I miss it? Man...that's pretty bad. I'll go make a flashcard now.

Megaqwerty Wrote:

snozle Wrote:I love the concept of an SRS in general, the fact that I keep acquiring material and don't have to worry about reviewing it, just let the software do it's thing.

Sorry, I looked over this thread, but I couldn't imply what SRS is. The concept of "not having to worry about reviewing, just let the software do it's thing" is certainly quite interesting.

EDIT: Oh, yeah...it hit me. Like KBAM, how could I miss it? Man...that's pretty bad. I'll go make a flashcard now.

Yeah I guess he really meant "not have to worry about scheduling my reviews, just let the software do it's thing"

Remember SRS's do not remove the need to study, they just stop you forgetting what you have already studied...

Yes, but we're splitting hairs. What he meant to say was that he doesn't have to go back and look at the stuff he's learned, because the SRS brings it right to him.

snozle Wrote:My question for you, who seem more experienced than me when it comes to SRS usage for Japanese, is what exactly do you do for vocabulary? Do you use individual words and translate them to English or do you do the "sentence method" like outlined at All Japanese All the Time? I kind of agree with khatzumoto, the author of All Japanese All the Time, when it comes to vocabulary in context; however, the problem that I have with his method is that he says not to start until after you have finished RTK1 because he encourages the sentences be from kanji to kana (japanese to japanese). That's great, but I feel that I should be doing something while I work through RTK.

Any suggestions?

I would really like to get in the position where I could use khatzumoto's technique effectively, but I'm maybe not yet there. I think the technique maybe does require finishing RTK1 to be really useful. If I fill my questions with kanji that I don't know yet then I will be working against RTK.

Without doing Kanji to Kana, It strikes me as more a technique for learning phrases and sentence patterns than vocab, but I could be wronng. Reviewing from Japanese sentences to English is passive review. By doing so you are more testing your reading & grammar skills than your vocab. Especially in sentences, the context can be a big crutch in recalling words. It may be much harder to recall the Japanese word when you need to use it later, or when you see or hear it in a different context.

Instead right now (while working on RTK1) I am mostly doing quite the opposite. Learning many words in isolation, and reviewing mostly English -> Japanese (looks down in shame). I do E->J for active recall which is quite important I feel.

The key thing for me is make the question as clear as possible. If its something like Q: "that" A"それ" there will be half a dozen ways to say it based on what you're talking about, how polite, etc. For that I often put the word inside a phraze, and also specify if its humble or honorific, for verbs - who's doing the action, for adjectives what is being descriped.

If the question is ambiguous then I will just recall the wrong word with similar meaning. Thats not the end of the world though, I will take the opportunity to add more detail to the question when that happens.

Something I started doing, which I'm really glad I got the idea to do, I bought a little white board and dry erase markers (the fine tip kind here http://expomarkers.com/sanford/consumer/...OPAT100003).

All the vocab I learn first goes on there (and RTK kanji too). They feel like you write with a little brush. These markers have a little eraser on the end that are very quick to correct errors, or blank stuff out for quizzing myself. Not only has it been great writing practice but I thoroughly quizz myself for active recall somewhere between 1 to 2 days before anything goes into Supermemo. I find that this makes more things stick first time and I have fewer lapses in Supermemo.

snozle Wrote:My question for you, who seem more experienced than me when it comes to SRS usage for Japanese, is what exactly do you do for vocabulary?

Any suggestions?

Program I use is Twinkle.

I've got various stacks of various styles.

Firstly, my kanji production practice (apart from this site) is based on practice for the Kanji Kentei tests. See the RTK2 Methodologies thread for my card styles.

Next, I have a JLPT grammar stack,

Q: example sentences with the "grammar nugget" blanked out.

A1: missing phrase
A2: meaning
A3: "grammar rules"

e.g.

Q: 最近テロ＿＿＿ニュースがよく新聞に出ている。

A1: に関する
A2: 〜について
A3: N１＋に関する＋N２

I've also got stacks just for reading/vocab practice. Some that I made myself and others with kanji tested in past JLPT1 and JLPT2 tests.

e.g.

Q: 案内
A: あんない

For cards that I make myself, I prefer to be familiar with the word but perhaps not the word/kanji relationship.

In the One question...Does it work? thread,

JimmySeal Wrote:Last October I scored 97% on Kanken Level 6, which requires being able to write 825 characters from memory and know their readings, compounds, and radicals. I also scored 70% on a sample test for level 5 (1006 characters) at the same time. Last December I passed JLPT level 1.

I'm taking the KanKen 6 tomorrow and am scoring 70-80% on past tests. So I'd like to hear how JimmySeal goes about things!

synewave Wrote:

JimmySeal Wrote:Last October I scored 97% on Kanken Level 6, which requires being able to write 825 characters from memory and know their readings, compounds, and radicals. I also scored 70% on a sample test for level 5 (1006 characters) at the same time. Last December I passed JLPT level 1.

I'm taking the KanKen 6 tomorrow and am scoring 70-80% on past tests. So I'd like to hear how JimmySeal goes about things!

Well, 8 years of studying on and off gave me a good head start, but for the test, I used the ステップ books meticulously. For each step, I would:

(1) Use the practice writing booklet.
(2) Look up all the example compounds I didn't know (this was before I stopped using dictionaries). For each compound, I would look up the word in my J->E dictionary, spend about 3 seconds looking at the entry, and then go on to the next one. I didn't write anything down.
(3) When I'd done that for all 6 or 7 characters, I'd do the exercises and grade myself, then look up the words I didn't know from the exercises.

I'd shoot for doing one step a day (finish in a month), and try to do parts (1) and (2) from above in three small bursts to keep from wearing out.

When I finished the ステップ book, I went through the 過去問題集 for level 6. I think this helped my score by 8-10 points. There are a lot of questions that pop up again and again. There's also a bit of luck involved. I was just lucky to get a test with only two questions I didn't know (one was 河原, a word that now mocks me every time I see it).
The one other thing I did was re-do all the 書き取り questions in the ステップ book, using that cute piece of cardboard they provide to cover up my old answers.


The only things I do now to study are:
Reading books and
studying 読み方. When I repeatedly come across a word I can't pronounce (and I'm near a table), I write it down on a slip of paper that I use as a bookmark. When I've gathered a bunch, I look up their readings and plug them into my flashcard program with Q: kanji and A: kana. I'm doing the same thing with Chinese now, but most of my Chinese flashcards are individual kanji, not compounds.

snozle Wrote:very interesting write-up vosimiura, thanks! I purchased a license to supermemo after reading rave reviews and I've never really used it. I only have macs at home and though I run windows in parallels and bootcamp I find it tedious to switch simply to review flashcards.

Any reason why you're not using iFlash on your Macs?

Serge Wrote:Any reason why you're not using iFlash on your Macs?

I actually have a iFlash on my mac and have used it in the past. I don't know why I've never stuck with it, but I kind of like the 1-5 scoring of answers that SM and Mnemosyne use. iFlash's interval learning method seems to be the closest to a Leitner approach, but I still like Mnemosyne's algorithm.

Are you using iFlash? If so, do you use the interval method and do you use it on the default values? The one drawback I see in mnemosyne is it's apparent lack of support for more than 2 sides, which may lead me back to iFlash in the future.

snozle Wrote:

Serge Wrote:Any reason why you're not using iFlash on your Macs?

Are you using iFlash? If so, do you use the interval method and do you use it on the default values? The one drawback I see in mnemosyne is it's apparent lack of support for more than 2 sides, which may lead me back to iFlash in the future.

I absolutely love iFlash - have tried some other software but always come back to iFlash. The design is great - very minimal and Apple-like. There is no limit to the number of cards or card sides; and you can even embed pictures and (I believe) sounds. You can change the font and the layout of the card, the only thing you can't change for some reason is colour - so I can't colour-code my Mandarin tones, which is irritating.

I always have it on Leitner repetition but I have changed the default value for incorrectly remembered cards: the default 'mistake' makes you lose 2 points and I've changed it to one, this works better for me.

There is also flexibility in reviewing a limited number of cards at a time or all of them at once. I have huge vocabulary and character files so I tend to review by level, starting from the highest. This also allows me to concentrate and prioritise better: e.g., if I'm reviewing something that has a value of 7 points (i.e., I have succesfully managed to recall it 7 times), I know I have to know this item very well, including (for Mandarin) the tones. If an item is at 1, I tend to be more lenient, if it's at 3 - I know I have managed to recall it properly before so this makes me think a bit harder. With lists running into (literally) thousands of items this methodolody seems to make sense.

I'm only running 3 stacks at the moment: Japanese vocabulary, with 4 sides (kanji, kana, a rough translation and a dictionary entry with examples); Mandarin vocabulary (words or sentences, pinyin, rough translation) and all Hanzi (around 6000 entries - characters, pinyin, rough translation and the index number to link back to the book I'm using). For the moment I'm working hard to create the Japanese vocabulary and the Hanzi files, but I may expand the number of stacks in the future to, e.g., the four-character sayings or lexical units picked from my reading, etc.

I only wish I could run iFlash on my Blackberry. :-(