overture2112 Wrote:

partner55083777 Wrote:I don't understand how this works....it's hard for me to believe that Wozniak could say "On average you roughly remember 95% of the material."

1) The 95% is from a 10% forgetting index.
For other forgetting index values,

Code:

retention = -(forgetting index)/ln(1-(forgetting index))

2) I'm pretty sure you agree with him but are misunderstanding the wording. The idea is 10% forgetting index implies 90% recall rate at the time cards are due. But right after you review, the time they are farthest from their due time, they should be at 100% (since you literally just answered them less than a moment ago). Your ability to recall will slowly fade from 100% to 100-forgetting index (ie 90% in this case) at the time the card is due. Thus if you average your recall rate over all those periods of time in between, you can express it with the formula above (which is 95% for a 10% FI)

So then what does the 95% represent? Does it represent the number of cards in your deck that you should currently be able to recall? For instance, if you were asked any random card from your deck today, you would be able to recall it 95% of the time?

This is the part that I don't understand. Let's say 99% of the cards in your deck are mature cards that are due far the future. If you were asked any of those cards today, shouldn't you have close to a 100% recall rate (way over 95%) despite the fact that your recall rate for your daily reviews is only 90%?

overture2112 Wrote:Your ability to recall will slowly fade from 100% to 100-forgetting index (ie 90% in this case) at the time the card is due. Thus if you average your recall rate over all those periods of time in between, you can express it with the formula above (which is 95% for a 10% FI)

Also, is this really how memory works? Is your "forgetting rate" constant? (I'm not sure I'm using the right vocab here, so I'll give an example...)

Let's say you have a 90% recall rate. You review a card today. It gets a 1-year interval. If you were to review it again today, you should remember it 100% of the time. If you were to review it again in 3 months, you would remember it 97.5% of the time. If you were to review it instead in 6 months, you would remember it 95% of the time. If you were to review it instead in 9 months you would remember it 92.5% of the time. If you were to review it in 1 year you would have your normal 90% recall rate.

Are you sure memory works like this? From my personal experience using Anki, I wanna say no, but it would be much better to have hard data.

I guess you could continue farther than one year as well. If you were to skip your 1 year review, and instead review it in 2 years, you would have an 80% recall rate. If you were to skip your 1 year review and review it in 5 years, you would have a 50% recall rate.

This really seems to go against my experience, but I figure you might be able to say that Anki's algorithm breaks down if you skip a review.


Edit: After thinking about it more, I realize you might be right. But can you explain why the rate of forgetting has to be constant (i.e. linear)? Why couldn't it be that other type of line (exponential or something)?

partner55083777 Wrote:So then what does the 95% represent? Does it represent the number of cards in your deck that you should currently be able to recall? For instance, if you were asked any random card from your deck today, you would be able to recall it 95% of the time?

No. It represents your chance to recall any individual fact at any given time throughout it's interval.

partner55083777 Wrote:This is the part that I don't understand. Let's say 99% of the cards in your deck are mature cards that are due far the future. If you were asked any of those cards today, shouldn't you have close to a 100% recall rate (way over 95%) despite the fact that your recall rate for your daily reviews is only 90%?

This is correct. If all your cards are near the start of their interval the recall rate is close to 100%. If we fast forward time to when they're all due, it'll be close to 90%. If you were to test yourself every second between (and magically had your memeory of these tests wiped) and averaged all the recall rates (which would vary between 100% and 90%, slowly falling as the time got closer and closer to the due date) it would average out to 95%.

partner55083777 Wrote:Also, is this really how memory works? Is your "forgetting rate" constant? (I'm not sure I'm using the right vocab here, so I'll give an example...)...But can you explain why the rate of forgetting has to be constant (i.e. linear)? Why couldn't it be that other type of line (exponential or something)?

The decay rate actually is exponential. This is perhaps hard to see with the 10% FI -> 95% retention, but observe these values:

Code:

Forgetting index    Retention
3%                  98.49%
5%                  97.47%
10%                 94.91%
15%                 92.29%
20%                 89.62%

Wozniak Wrote:The reason that the retention is not equal to 1-0.5*(forgetting index) is that forgetting is approximately exponential in nature. Immediately after the repetition, forgetting proceeds at the highest rate.

Source

overture2112 Wrote:More accurately:

By "retention" I assume you mean "forgetting index", which are different, as generally retention is considered to be your average ability to recall at any given time.

Wozniak Wrote:If you set your forgetting index to 10%, you will remember 90% of the material at repetitions. This does not imply that your knowledge retention will be 90% only. Your average retention will be nearly 95%! This comes from the fact that 90% refers to the retention at repetitions, while the initial retention right after the repetition is theoretically 100%. During the inter-repetition interval, retention is decreasing from 100% to 90%. On average you roughly remember 95% of the material.

For optimal forgetting index: "Simulation experiments have consistently pointed to the value of 25-30%." But he recommends no higher than 20% because generally most facts have some interdependancies and there's also a negative effect wrt esteem/motivation/stress from forgetting too much.

Source

Yup, that's what I was thinking of. Thanks for the clarification and correction!

nadiatims Wrote:In the day and age of pop-up and electronic dictionaries it seems rather pointless to me to SRS vocabulary. Just keep reading and if you've forgotten a word, just look it up again. It's no biggie.

You gain way more benefit from reading real sentences within the context of real articles or stories than from just flash carding vocabulary, because it also teaches you grammar and gets you thinking in the target language for longer chunks of time. Basically it's more of a linguistic workout.

For some of us reading material we don't understand is not that fun of an activity. Especially for those that really need to stop and think about the word to be able to memorize it. It breaks the pace completely and I've completely forgotten what the rest of the sentence was about in the meantime.

I find it much more enjoyable to keep reading while glancing at the words I don't know, focusing on the translation and then moving on. The big picture stays clear and I don't get stuck on a word for too long. We will meet again in Anki in where the focus is completely on the word itself.

Sure, once it appears that I only need to look up every 10th-20th word it makes more sense to try to memorize it then and there. That said SRS will take 10 minutes of your time at that point and you get to see rarer words more often, I don't see many negatives about this.

^ plus

There is this other aspect of using SRS alongside reading exercises - the discipline. I don't know about others but knowing that there are due cards waiting for me makes me do those reviews and stay in touch with the language. If this was for reading only I might feel sometimes like not doing it and that would be a problem.

I view the combination of reading (or any other form of immersion) and SRS as an “offence on the language” and “defense against forgetting” respectively.

overture2112 Wrote:The decay rate actually is exponential.

Okay, I think I kind of understand. Thanks.

Here is also another interesting page linked to from the source above:

http://supermemo.com/articles/theory.htm

Edit: However, I don't some of it. Do you know what he means by "A-factor"? Is it just the difficulty of a certain card?

Edit 2: A couple of interesting things taken from the page above:

Wozniak Wrote:・In a long-term process, for the forgetting index equal to 10%, the average rate of learning for generic material can be approximated to 200-300 items/year/min, i.e. one minute of learning per day results in acquisition of 200-300 items per year.
・Eliminating 10% of the most difficult items in a generic material may produce an increase in the speed of learning of up to 300%. The lower the forgetting index, the greater the increase.
・In a long-term process, for the forgetting index equal to 10%, and for a fixed daily working time, the average time spent on memorizing new items is only 5% of the total time spent on repetitions. This value is almost independent of the size of the learning material
・In a long-term learning process, 50% of repetitions are devoted to 2.5% of short-interval learning material (actual learning process measurements). This number can vary greatly in practice and in ill-structured learning material, even a smaller proportion of items can take most of the learning time.

I currently have the 'Suspend Leeches' functionality disabled in Anki, but it might be a good idea to re-enable it!

Edit3: It would be nice to be able to see a lot more of these statistics in Anki. For instance, if I wanted to turn on "Suspend Leeches", it would be nice to be able to see a graph that could tell me how many failures I should set the leech threshold at. For instance, it could show me a graph where I could tell that if I set the leech threshold at 10 lapses, then Anki would end up suspending the hardest 8% of cards, which would give me a 150% increase in efficiency. That would be really cool.

Betelgeuzah Wrote:

Quote:A thing that people should keep in mind is that once your level of vocabulary reaches a certain point, you no longer really need to study vocabulary in order to learn more of it. I believe the transition point is somewhere in the neighborhood of 15,000-20,000 words.

I'd say even that many words isn't a necessity. My English vocab is somewhere around 12,000 known words and probably half of it was learnt "in context". I was lazy in school but active in the interwebs.

More important is probably actually getting used to the language, not necessarily the amount of vocab one learns. There are so many words that I "know" but can't translate to Finnish or explicitly define them without looking at my English dictionary. Many times a word pops in my head and I am left wondering "how the hell do I even know this word and how to use it?"

Perhaps Japanese is different...but certainly the kanji should help one decipher the meaning.

I totally agree. Anki itself not get you to fluency. Plus not all the words you know or words you can infer the meaning are in your anki deck. stop doing silly math. spend that time doing something in japanese DUH. I was thinking that one thing that Japanese children probably till have a one-up on me is gi-on and that's precisely because they're more used to japanese than me (since they only know Japanese and it's been since birth for however many years). the thing with gi-on is, you gotta feel it... it just sounds like what it is.

Zgarbas Wrote:Personally, I can count all words I actually learned in native media on my fingers. I'll encounter them, look them up, and instantly forget them*. Even when I mine them into an anki deck, my retention rate is pretty much the same as with words in premade decks. Same goes for words encountered in sentences; I have some sentences which I keep failing because I don't remember a word which I encountered there for the first time... they just don't stick to me.
I'd much rather spend an hour in anki learning&reviewing hundreds of terms and sentences than spend 20 hours with native material, and anki is more efficient. Learning them in anki, then encountering them in native media to reinforce it is what works best for me.

*unless they're ridiculously common; see all the anime terms that even people with no interest in Japanese automatically learned. Titles or catchphrases also work sometimes. They're usually basic words, and the amount of exposure necessary to actually learn them is pretty time-consuming.

(sooo, each with their own, eh?)

I have the completely opposite experience. a lot of words and phrases I learned from native media were so effortlessly learned. I learned it into context which was very memorable or funny or something. I found talk/variety shows really useful and fun for learning common vocabulary. It's simple and common vocabulary YET it's funny and entertaining. After a while I realize they just use the same words over and over for the most part since it is DAILY conversation. I saw the core6000 deck and it's like nothing. all the words are used a lot in everyday conversation so i don't see the need to torture yourself with that boring deck especially with all the talk/variety shows available online or the subs to dramas/anime/etc or music lyrics even. core6000 is tip of the iceberg for fluency. I think you just haven't found any media that really pulls you in....

The only time I would have the same reaction you would is if i was really beginner and that stuff was too hard or i was trying to do too much or watching something really crappy and boring. but i started watching talk/variety after tae kim so it's not like i was even that good at the time.

^How are are you watching variety shows. Are there torrents.

[s0apgun]

^^ This post reeks of AJATT regurgitation.

s0apgun Wrote:^^ This post reeks of AJATT regurgitation.

Whatever works, works. Does it matter where it comes from?

s0apgun Wrote:^^ This post reeks of AJATT regurgitation.

this post is needlessly antagonistic towards someone with a different learning methodology.
besides is regurgitation really such a crime? this is a language learning board not a creative writing class... I'd say you regurgitate a fair amount too otherwise you should write a book about all your unique ideas.

[s0apgun]

Putting words in my mouth but whatever. Some people on this forum who keep pushing towards immersion and media is better don't quite seem to get that it requires a whole lifestyle change to full benefit from that. People using anki use it for its efficiency and don't find it boring it all however they keep INSISTING drilling Core2k is for robots.

I barely have time to have a couple hours everyday to myself to relax, let alone going through media I can't comprehend at all yet and looking up words in a dictionary (which isn't as easy for Japanese as everyone makes it out to be unless you're on a computer). Obviously, both methods work but everyone gets all worked up on here because they've become FANS of this or that because of things they read on the internet. That being said, my method is Anki and audio as much as I can. It works great for me, probably not for others!

Betelgeuzah Wrote:

nadiatims Wrote:In the day and age of pop-up and electronic dictionaries it seems rather pointless to me to SRS vocabulary. Just keep reading and if you've forgotten a word, just look it up again. It's no biggie.

You gain way more benefit from reading real sentences within the context of real articles or stories than from just flash carding vocabulary, because it also teaches you grammar and gets you thinking in the target language for longer chunks of time. Basically it's more of a linguistic workout.

For some of us reading material we don't understand is not that fun of an activity. Especially for those that really need to stop and think about the word to be able to memorize it. It breaks the pace completely and I've completely forgotten what the rest of the sentence was about in the meantime.

I find it much more enjoyable to keep reading while glancing at the words I don't know, focusing on the translation and then moving on. The big picture stays clear and I don't get stuck on a word for too long. We will meet again in Anki in where the focus is completely on the word itself.

fair enough. It doesn't really matter when or where you learn words, your reading ability should improve as you improve your vocabulary. Actually my criticism of SRS is less relevant for people who rely on it less. Counter intuitively I think it's more effective the less you use it. I think for people spending much more than say 10% of their study time in the SRS there is high potential for inefficiency and some ineffectiveness.

I've been beginning to wonder recently whether I'm the only person who thinks this way...So I quickly googled it. Here's what tae kim has to say. Particularly the following rings true to me:

"Second, the whole review thing seems backwards to me. If I review a word that’s completely new, I pick “Hard” and then it shows up again right away. For me, seeing a word I don’t know over and over again does not help me. I need new words to bake over time. If I know the word, I want to delete it, and if I don’t know it at all, I pick “Easy”. If the word looks familiar to me, I pick “Hard” so that I can see if I want to delete it the next time."

and

"Don’t sit around wasting time entering and reviewing what you’ve already seen, just get more, more, and MORE STUFF!!! You’ll be surprised at how much just seems to stick somehow like osmosis. Some people feel this is not effective because they end up forgetting so much stuff. They don’t realize that the fact that they even remember forgetting it means they’re learning it."

nadiatims Wrote:

Betelgeuzah Wrote:

nadiatims Wrote:In the day and age of pop-up and electronic dictionaries it seems rather pointless to me to SRS vocabulary. Just keep reading and if you've forgotten a word, just look it up again. It's no biggie.

You gain way more benefit from reading real sentences within the context of real articles or stories than from just flash carding vocabulary, because it also teaches you grammar and gets you thinking in the target language for longer chunks of time. Basically it's more of a linguistic workout.

For some of us reading material we don't understand is not that fun of an activity. Especially for those that really need to stop and think about the word to be able to memorize it. It breaks the pace completely and I've completely forgotten what the rest of the sentence was about in the meantime.

I find it much more enjoyable to keep reading while glancing at the words I don't know, focusing on the translation and then moving on. The big picture stays clear and I don't get stuck on a word for too long. We will meet again in Anki in where the focus is completely on the word itself.

fair enough. It doesn't really matter when or where you learn words, your reading ability should improve as you improve your vocabulary. Actually my criticism of SRS is less relevant for people who rely on it less. Counter intuitively I think it's more effective the less you use it. I think for people spending much more than say 10% of their study time in the SRS there is high potential for inefficiency and some ineffectiveness.

I've been beginning to wonder recently whether I'm the only person who thinks this way...So I quickly googled it. Here's what tae kim has to say. Particularly the following rings true to me:

"Second, the whole review thing seems backwards to me. If I review a word that’s completely new, I pick “Hard” and then it shows up again right away. For me, seeing a word I don’t know over and over again does not help me. I need new words to bake over time. If I know the word, I want to delete it, and if I don’t know it at all, I pick “Easy”. If the word looks familiar to me, I pick “Hard” so that I can see if I want to delete it the next time."

and

"Don’t sit around wasting time entering and reviewing what you’ve already seen, just get more, more, and MORE STUFF!!! You’ll be surprised at how much just seems to stick somehow like osmosis. Some people feel this is not effective because they end up forgetting so much stuff. They don’t realize that the fact that they even remember forgetting it means they’re learning it."

You're not the only one who feels this way. I feel like Anki is more useful for me the less I use it. When I was spending 1-3 hours a day with it, the words absolutely did not stick. However, using Anki for around 1/2 an hour, and using the rest of my time with a textbook or reading a novel, writing down unknown words, and looking up J-J or J-E definitions and writing those down in a notebook works so much better. I noticed that a lot of the words I had to look up were words that I had seen multiple times in Anki. Just SRSing doesn't work for me. I'm starting to view SRS as more of a backup method that I can ignore when I want, whereas my textbook and novels are my main source of study.

nadiatims Wrote:"Don’t sit around wasting time entering and reviewing what you’ve already seen, just get more, more, and MORE STUFF!!! You’ll be surprised at how much just seems to stick somehow like osmosis. Some people feel this is not effective because they end up forgetting so much stuff. They don’t realize that the fact that they even remember forgetting it means they’re learning it."

This is exactly how I felt about RTK, while I was going through it at a fast rate (fortnight to finish reviewing them all, etc.). It seemed that the ones I would forget earlier, for whatever reasons, no longer become an issue as I continued to add more and more cards to the daily review pile. It was almost like an overload wherein while my initial retention rate may not been as high other others but the vast amount of input I was able to identify with being something I forgot, as opposed to something entirely alien, was a strange experience.

Even now, having stupidly taken time off reviewing Core vocab, often come across kanji which I get a feeling for the reading, or the meaning, or how it relates to another word family, just by lazily reading various chunks of text in various manga *cough*hentai*cough and other Japanese, along with listening to Japanese music, Japanese animation and movies, etc. It's almost as though the more you stuff your face with, the more sense patterns make, how formally studied things work, etc.

corry Wrote:^How are are you watching variety shows. Are there torrents.

yes but I mostly watch off youku. U just google douga + show title. Nowadays the video quality is high

s0apgun Wrote:I barely have time to have a couple hours everyday to myself to relax, let alone going through media I can't comprehend at all yet and looking up words in a dictionary (which isn't as easy for Japanese as everyone makes it out to be unless you're on a computer). Obviously, both methods work but everyone gets all worked up on here because they've become FANS of this or that because of things they read on the internet. That being said, my method is Anki and audio as much as I can. It works great for me, probably not for others!

So you're completely avoiding native media until you can completely comprehend it? It's understandable if you are working on core2k, but after that, you should be able to read most "easy" stuff without too many issues. Heck, I've only got 2800 (self-input) anki cards, and can read Kino no Tabi without major problems understanding it.

this site has good jTV too:
http://moviemaga.blog20.fc2.com/

variety/dorama etc.

Also re anki, I think the inefficiency part really kicks in when it comes to entering cards. I know we all like to complain about how boring core is and how subs2srs cards can be too hard... but the dark side to that pretty entertaining deck with every card custom-made from your favorite sources is the enormous time it took to make it. When I started out doing sentences I spent several hours every day for a year just making cards! It was at least half my study time!
Now I'm into mass-generated/pre-made decks or nothing. I have another 1,000 left in core10k and after that it's subs2srs until I stop adding I think.

Second thought I might do a people/place names deck... so many things to learn, so little time.

Huh? I use about 10-15 minutes per 10 cards... though I'm mostly using subs2srs (and additional cards for "j-j branching" nowadays.) otherwise it might be 15-20. So Anki activities totals about 45-60 min (including reviews)

Either you were just very slow, or you added way too many cards. Well, it might only be you exaggarating to make your point valid.

Subs2srs is a lot of work as well; of about 300 cards, you would have to delete about 250*, and you would have to find definitions for the rest. Finding Japanese subtitles isn't the easiest thing in the world either.

*Even more if you have a vocabulary of 10k words, I assume.


EDIT: Oh, and thanks a lot for that website; I've been looking for a RAW stream site I started out last summer. :p

Maybe we should just increase the Interval Modifier in Anki? I am going to try 400%.

Stian Wrote:It's understandable if you are working on core2k, but after that, you should be able to read most "easy" stuff without too many issues.

... or listen to relatively simple speech - like this one:
杏のanytime andante #2

Hint: she's talking about travelling.

corry Wrote:Maybe we should just increase the Interval Modifier in Anki? I am going to try 400%.

This is a joke, right?

What is your current retention%? Is your desired retention like 0.007%?

log(desired retention%) / log(current retention%)

corry Wrote:Maybe we should just increase the Interval Modifier in Anki? I am going to try 400%.

if interval modifier does what it sounds like it does, I'd recommend setting it to something like 2000%.