Spaced Repetition Notes System (SRNS)

[juman]

Spaced Repetition Notes System (SRNS)

You have probably heard about SRS applications like Anki, Memrise etc and even though they are good after spending too many hours per day in front of the screen I want something getting me AFK (away from keyboard). So here I will present my ideas, views, trials, errors and hopefully a success in the end in creating a paper based system that can be part of my process learning a new language.

But there is already a bunch out there you say why not use one of them? The answer is they do not provide what I believe is a good solution. So here is what I am hoping to be able to have as features when finding a process for this.

Feature goals :

• Self contained - in notebook/s or within loose leafs
• Portable - I travel so I want to be able to bring it with me
• Without computer need - Reviews shouldn't need a computer
• Phase out - Have a definition of when something is learnt
• Gamified - Possibility to track things added/learnt
• Handwritten - I feel that writing things out by hand helps a lot in the learning process
• Simple - Simple enough to manage that the effort to use it is low

If I look at the current systems I know of we have the following :

Flash cards (https://en.wikipedia.org/wiki/Flashcard)
Could be useful but I'm imagining that the system needs to be able to handle at least 500-1000 words/phrases. That would be quite a pile of index cards and far too much to drag around.
Leitner System (https://en.wikipedia.org/wiki/Leitner_system)
Flash cards based so having the issue mentioned above. I like though that there is a defenition of when something is done and moved out of the system.
Paper-and-pencil SuperMemo (https://www.supermemo.com/articles/paper.htm)
Could be something but seems complex in the layout between two notebooks and the description states that it is too much overhead and that is why the software is needed.

Next up will be a description on my first attempt to find a solution.

[rdearman]

I have printed ten sentences on a single index card, with answers on back. This means you only need 50 cards for 500 sentences, or basically one deck.

Another thing I have done in the past is create 8-page booklets on a single sheet of paper. You can print anything you want per page and it is scaled down to only 7cm x 10cm 8 page booklet you can fit into a shirt pocket. This would allow you to have up to 20-25 lines of readable text on each page, so you could potentially have 100 flashcards per booklet, so 5 of them would do 500 sentences. Instructions for this below and a template for OpenOffice (There are Word versions too)



Template: https://templates.openoffice.org/en/tem ... ge-booklet

The benefit of this booklet is you can print anything you want on to it. It is just a small 8 page document turned into a booklet.

As far as being done with flashcards there is a system I learned long ago. Each time you review the card if you know it you put a o on the card. Each time you get it wrong you put a small x. You retire the card when you have 3 consecutive o 's

[juman]

Thanks for the reply on the flashcards rdearman and Hashimi. Having multiple words on a flashcard is something I do at times for extra drilling however it leaves out some things I'm thinking about how to address :

Say you have the following front of a card :

dog
cat
house
gross domestic product

Three of the four phrases will probably be quite easy but the fourth is a bit trickier. So how do you value the card in overall difficulty? Do you repeat in until you know all words or do you move the difficult word to a card with only trickier words which you drill more frequently? Another challenge for me and flash cards is sentences as I like to create them by hand. A sentence can easily fill a whole flashcard so once again back to the sizing issue.

[juman]

I spent some time reading through https://www.supermemo.com/articles/paper.htm to get a better understanding of what it is suggesting that you do. Summarized it describes the following steps :

1. Write down a list of Q&As
2. Test yourself on the Q&As in order
3. Put a dot on the failed Q&As
4. Go through the dotted Q&As and repeat the procedure until you know it all
5. Then follows a kind of scoring of the results I don't really get and it doesn't seem to impact the interval until next repetition
6. Next repetition date is taken from the static list of intervals

So my issue with this is that it only seems to use the static list of intervals to choose when the next iteration will be and looking into more modern SRSs they adjust dependent on the difficulty of what you are learning. Some things needs to be reviewed more frequently then others which this system doesn't seem to handle. Also I don't understand the use of the scoring and if it actually has any impact at all.

[rdearman]

Personally I wouldn't bother doing all of that. I would simply download anki load and some cards and start using it to memorise vocabulary. All of that other stuff just seems to be procrastination to me.

If you're determined to use paper than you might want to read Iverson greensheet method or look for the gold list method.

[juman]

If you're determined to use paper than you might want to read Iverson greensheet method or look for the gold list method.

I have seen Iversons method of translating back and forth which but the green sheets where more of creating a reference sheet as I understood it?

Regarding the gold list method it has one good concept as it keeps you writing out the material. However when it comes to the actual structure and baseline for the technique it sounds like a dream. However a dream is rarely reality and that is my experience with the gold list method. In short I have tried it and the results where not good.

[juman]

I dived in the the documentations I could find regarding what interval to use when it comes to SRS and there seems to be two options. One way is to use a static list of intervals as Leitner and Supermemo on paper. Alternative number two is to use an algorithm which increases/decreases the interval based on how easy you thought the recall was. Anki is using an algorithm called SM-2 (https://www.supermemo.com/english/ol/sm2.htm) which I draw as a flowchart in the following diagram :

sm-2.png

The algorithm is not that complicated and made me learn some new stuff about Anki. If you say that the ease is 0,1 or 2 the result is the same and you will see the card again the next day. If you vote 4 you are not changing the Ease-Factor and the increase of the interval will be static.

Now on to read up more on algorithms of today and also to see if this can be simplified somehow to be used by hand. Calculating using SM-2 will require a calculator as I am not that good at math so I can't calculate EF - 0.8 + 0.28 * q - 0.02 * q * q easily.

[juman]

I read up some more about the latest algorithms from Supermemo and in short they seem to use a larger dataset to set the next iteration per item. In short it seems far to complicated to apply with pen and paper. So I looked through the ones that seemed easier and compared them to the Fibonacci scale (only reference I could think of).

In the picture below I have Leitner, SM-0 (Supermemo with pen and paper), SM-2 where you only choose 3 or 4 or 5 as your answers. I thought they would be quite similar but the increases of the steps in SM-2 is much higher then the Leitner system. So if you go 7 iterations you would have an item in your system the following amount of days (with the longest interval in paranthesis) :

Fibonacci = 33 (13)
Leitner = 127 (64)
SM-0 = 223 (90)
SM-2(3) = 473 (228)
SM-2(4) = 973 (586)
SM-2(5) = 1550 (1026)

srs.png

This makes me think that using this with pen and paper I probably want to cut off items off at some point and see them as learned. If I then run in to an item again in the future and have forgot it I can reenter it again. Keeping track of items to repeat them again in 1026 days (2.8 years) just seems like overkill. In that time I hopefully will have run into it again or otherwise it is probably not relevant.

Reading about Leitner it seems common that after you have remembered an item for 64 days you see it as learnt and move it out of the system. With that cut off timespan the different systems would have the following amount of iterations when it is easy to learn :

Fibonacci = 10
Leitner = 7
SM-0 = 6
SM-2(3) = 5
SM-2(4) = 5 (will overspan the 64 days with 34 days in the last iteration)
SM-2(5) = 5 (will overspan the 64 days with 53 days in the last iteration)

[juman]

So I considered how to use the spaced repetition system that Anki uses but with pen and paper and if that was even possible. For each item to be memorized two states need to be kept between sessions. One is the last EF-number and the other one is the last timespan since the previous review.

So using these two tables the repetitions could be done like this :

NewEF.PNG

new Iteration.PNG

1. Write your item down
2. Day 2 review the item and grade how easy it was to recall
3. In this example lets say we thought it was hard to recall so get the EF value of 2.3 from the bold line
4. In the second table follow the 2.3 column to the 1 day row and you get 2 so schedule a review in two days
5. Two days later we review the item again but and still think it is hard to recall
6. In the cross section of Hard and 2.3 in the first table we get 2.1
7. In the second table in the cross section of 2.1 and 2 you find 4 so now schedule the next review in 4 days

[zenmonkey]

why is "again" and "good" the same?