Wednesday, April 24, 2013

Finally Posting Again: Small Timer = Huge Success

I haven't been posting anything for quite some time for a number of reasons, but it all comes down to trying to strike a balance with posting things and keeping up with the thousands of other things I want to keep up with.

This is the main reason, though: for the past couple of months, I have been testing the use of a timer to "prime" memories for long-term retention by reviewing new material after 1 minute, 3 minutes, 10 minutes and 1 hour. It has been very successful at helping me commit very new things to memory. This includes:

-New Chinese characters
-Words in an unfamiliar foreign language for which I don't have an immediate frame of reference (Hindi, Russian, etc.)
-Location based mnemonic placeholders (Method of Loci) (More on this later)
-Dates and other number-based information

It has also been very helpful at committing to memory things that I already have an "edge" with, but it ensures that I almost always recall it after the first interval in SuperMemo. For example, in the past I would often learn a new Japanese or Chinese word, put that word into SuperMemo, but I couldn't quickly recall it when I had to review it for the first time in SuperMemo, often 4-9 days later. I would get the item wrong, then maybe get it wrong once more (2-3 days later), and after that I usually had no trouble remembering it. Since I began using this timer method for introducing material, very little has been forgotten when I review it for the first time. Words that I learn can be used almost immediately, which wasn't always the case in the past.

Since my previous post, I have added about 5,000 new flashcards to SuperMemo using the timer method to introduce and acquaint myself with the material. I have also been incrementally reading and learning various things related to science, art, religion, history, etc. (Which doesn't require using a timer to commit), but the bulk of my "experiment" time has been devoted to testing my capacity at learning difficult material using the timer.


Size of flashcard sets
I learned anywhere from 10 flashcards to 100 flashcards at a time (I went as high as 120 once or twice), and I found anything beyond 30 or 40 to be cumbersome, overly taxing and caused me to dread the review process. The "sweet spot" (For me, at least) seems to be 25-30 flashcards per set of new material. If the information is extremely difficult, 10-15 flashcards seems to be better.

Makeup of material
I found it helpful to mix extremely difficult material with slightly easier material. I'm not sure if this is for my own motivational benefit, but it was helpful at keeping me motivated. For example, one set of new material could consist of: 10 new Japanese words, 5 mnemonic items, 5 Chinese characters and 5 Hindi words. The mnemonic items and Japanese words are almost always easy to recall, while the new Chinese characters and Hindi words take a bit more effort. The smaller "easy victories" made it easier to exert myself at learning newer, unfamiliar stuff.

"Stickiness" factor
Since I began using a timer to commit "unsticky" information to memory,  it has almost always become "sticky" and therefore easy to remember in the short and long-term. A few items would be forgotten, but of the 5,000 or so items that I have learned over the past couple of months, a VERY small percentage of them have been forgotten (Less than 2%). In the case of those items, by simply "relearning" them using a timer to space out the short-term review intervals I've been able to "make sticky" those items also.

How I've integrated it
Whenever I have about 1 hour and 15 minutes of time, I start by reviewing the 25 new flashcards once or twice, then I start the first timer. After this, I surf the internet for a minute until the timer goes off; I then review the flashcards again and set the second timer for three minutes. During this time I typically do a small task (Clean up the immediate area, read a book, etc). After the three minute timer goes off, I review the flashcards and set a timer for 10 minutes. At this point I have quite a bit of time to do various things (Chores, prepare simple food, watch a TV show or news, incremental reading, etc.). After the 10 minute timer is done I review the cards again and set the timer for one hour. During this time, I can work on any extensive or engrossing task, and I often forget about doing flashcards until the timer goes off again. If necessary I can start learning another set of information during this hour, but I've enjoyed not doing that and chilling out a bit.

Thus, by adding just a bit of structure, I've found that I can quite easily integrate learning new, "unsticky" material while going about my day to day activities (Or at least the down time I have) so that I'm not spending 100% of my time on flashcards, but I still feel like I'm using my time wisely. As long as I have prepared material to learn, I can easily do 4 or 5 sets on a normal work day, which nets at least 100 new cards per day.

Implications
Having found a balance, I have been able to more precisely control the flow of new material into SuperMemo without overloading myself or creating a glut of incorrect responses. (For me) The obvious implication is that committing new languages to memory can become a very streamlined, steady and less painful process than it usually is. The same goes with things such as programming, medical information, etc., or careers that require committing "unsticky" frameworks of knowledge to memory. I'll elaborate on the language aspect in a future post, it's a different subject altogether.

7 comments:

  1. Glad to see you're posting again! Your superlong AJATT comment is what got me using SRS (in my case, Anki) for my college classes and I can tell a difference.

    Anyway, ever since your post on the Pimsleur microintervals I've been wondering how it'd affect vocabulary acquisition via SRS cards. So I set a couple decks (Spanish and French vocabulary, my target languages) with the "steps" being equivalent to Pimsleur's intervals. While it worked (after a while I changed to a simper format-10 mins, 1hr, then 1d, and this wasn't nearly as good), I wasn't able to get to them EXACTLY 2 mins, 10 mins, etc after I reviewed them most of the time, which wasn't optimal, and of course being in college and having classes and other crap going on I sometimes found it hard to actually get the longer intervals (e.g. I'd end up doing one that was supposed to be reviewed after 5 hours maybe 12 hours later the next day).
    Another factor that didn't help was that I think I learn vocab better in context, even via SRS. I ended up deleting those decks and have discontinued using microintervals for any decks (for practical reasons) but if there's someone who can make it work time-wise and can learn from isolated words (or I suppose you could do this with sentences), then it's definitely worth it.

    I won't be able to restart doing the microintervals this summer either, as I have an internship that'll probably be like a typical 9-5. But any other future ideas you come up with I'll probably jump all over and be a guinea pig.

    ReplyDelete
  2. Hi!

    This sounds really interesting. I myself I am working in something similar but with a bit different time pattern. If someday I have an space similar to yours I would love to post and share the results.... Anyways, there is a point I don't understand; hope you can make it clear to me. If every time the timer sounds you review all the flashcards at the same time straight, it doesn't mean that you are going against the "minium information" principle stated by Wozniak.

    I don't take Wozniak's words as a dogma, it's just that I am curious about how do you set out this point.

    ReplyDelete
  3. Glad to see you're posting again! Your superlong AJATT comment is what got me using SRS (in my case, Anki) for my college classes and I can tell a difference.

    Anyway, ever since your post on the Pimsleur microintervals I've been wondering how it'd affect vocabulary acquisition via SRS cards. So I set a couple decks (Spanish and French vocabulary, my target languages) with the "steps" being equivalent to Pimsleur's intervals. While it worked (after a while I changed to a simper format-10 mins, 1hr, then 1d, and this wasn't nearly as good), I wasn't able to get to them EXACTLY 2 mins, 10 mins, etc after I reviewed them most of the time, which wasn't optimal, and of course being in college and having classes and other crap going on I sometimes found it hard to actually get the longer intervals (e.g. I'd end up doing one that was supposed to be reviewed after 5 hours maybe 12 hours later the next day).
    Another factor that didn't help was that I think I learn vocab better in context, even via SRS. I ended up deleting those decks and have discontinued using microintervals for any decks (for practical reasons) but if there's someone who can make it work time-wise and can learn from isolated words (or I suppose you could do this with sentences), then it's definitely worth it.

    I won't be able to restart doing the microintervals this summer either, as I have an internship that'll probably be like a typical 9-5. But any other future ideas you come up with I'll probably jump all over and be a guinea pig.

    ReplyDelete
  4. Thank you for writing about your experiences.

    I've tried learning the greek alphabet using the time intervals you have described and it worked like a charm!

    ReplyDelete
  5. Thank you, LittleFish! I am planning to learn a Major-System and started experimenting with a timer. Your repetition cycle seems to be much more effective then my experiments. Great Post!

    ReplyDelete
  6. How do you usually add questions and answers to SuperMemo?

    I used to add them manually at the program, one by one. Then I learned that you can import a .txt file. So, now, I 'found' a way to make this faster:

    You make a template at excel, fill it up, copy all cells, paste into notepad, import it to SM.

    I managed to include a list with 1500 french verbs in just two days! It is insane, you should try if you haven't yet...

    In excel it looks like this

    Line ONE:

    Column A (in this column it is either Q:[question] or A:[answer]): Q:

    Column B ~OPTIONAL~. (this is where you determine via HTML codes how you want the question text to look, in case you don't want supermemo's default).

    Column C (this is where you put the word, question, or whatever you want to question about): QUESTION TEXT

    Column D ~optional~ (this closes 'line one' as well as HTML tags)


    Line TWO:

    Column A: A:

    Column B: whatever HTML tags to make your 'answer text' look nicer

    Column C: ANSWER TEXT

    Column D: closer html tags


    Line THREE

    leave blank!

    -
    Line one is QUESTION. Line two is ANSWER. Line three is how SuperMemo separate one element from the other. With excel you may copy this 'template' in the spreadsheet by just selecting all of it (including Line Three) and pushing the selection downwards.

    Let's say you pushed down to create 150 'copies'. Now, all you have to do is fill up Column C, question and answers. Copy the cells to notepad. Save. Make supermemo import text. Et voilà ! 150 new items pretty fast.

    This is much, much, faster! You could do that with Notepad, simply, but editing at excel makes it faster and easier...

    if you want more explanations or a copy my templates, email me at egoals2 at gmail dot com

    :) [ really like your blog ]

    ReplyDelete
  7. This is a change from the expanding repetitions that you derived from your Pimsleur audio tracks. It has fewer repetitions and it eliminates that long 5-hour delay.

    This schedule is more efficient. Do you find it as effective for retention?

    ReplyDelete