How to best remember new words

@Achab:

>>"[Fullrecall and MemAid] calculate the next optimum point of time when to best review new vocabulary".

>>This is the same way that Supermemo works


As mentioned earlier there is one but crucial difference between FullRecall/MemAid and Supermemo:

Supermemo's way of calculating the next optimum time for repetition is fixed - no matter whether you are a "good learner" or a "bad learner".

FullRecall/MemAid's settings however also take the individual abilities into account which is why the intervalls are not "hard-wired" but can change.

Let's take a simplified and hypothetical example (the dates shown here are just for illustration purposes and might of course be totally different in reality):

1. Supermemo's calculated days for the next repetition intervalls for a "very easy" word, i.e. in case the learner remembered a word perfectly well each time, starting with day 1 = first time when learnt a new word:

1 - 8 - 20 - 34 - 87 - 210

2. FullRecall/MemAid's calculated days for the next repetition intervalls for a "very easy" word, i.e. in case the learner remembered a word perfectly well each time, starting with day 1 (= first time when learnt a new word):

1 - 8 - 20 - 38 - 110 - 365

In this case these programmes might have found out after the third feedback that this particular learner seems to be a good learner who's last optimum date for repetition might therefore be in 365 days time and not after 210 days as with Supermemo.



I hope this has made the differences and levels of efficiencies between the relevant programmes clearer.

We are discussing different ways of learing vocab and I have explained the basic methods we use.I think this can be of interest. If Tom cuts me off so be it. I think there is value for all in these discussions.

Boy,

We ask the learners to assess themselves, this limits the goals they can set for themselves. If they over or under assess themselves they will soon see that they are unable to keep up to their goals or surpass them too easily. Then they can go in and adjust their goals. I do not believe too much in objective assessments. I think most people know how they are doing with the language. They know that they need to learn more. Where they are on the ladder is not so important in my view. Once in the system they have numbers and graphs showing their own progress.

The goals you show are not impossible to achieve, but if you think them too difficult you just adjust the time to 12 months rather than 6. Remember that you will gradually get credit for words you already know and this will help you achieve your word goals. The listening from your list is a little heavy but the writing (300 words per week) is not too bad.

Achab,

I am fully aware of that fact.

DaVinci,

sorry, I really do not understand your explanation of how FullRecall works in a more intelligent way than Supermemo. What I guess from what you wrote is that FullRecall has a pattern of intervalls (1 - 8 - 20 - 38 - 110 - 365) that is different from the one used by Supermemo (1 - 8 - 20 - 34 - 87 - 210). I cannot get why FullRecall is more intelligent than Supermemo. Maybe the reason why I do not understand is that I have not used Supermemo to learn vocabulary yet. I am going to start soon though; these days I am making a lot of research on Supermemo and similar pieces of software.

Tom,

if you are out there, do you understand better what DaVinci said? Have you ever used FullRecall/MemAid? Do you have an opinion? (I am asking Tom because of his being knowledgeable about Supermemo-type technology.)


Achab

just do it.

people, stop beating around the bush and just learn your crap. no better way than the old fashioned way. sit down with a book, memorize, read, learn, enjoy, and if you do this for at least 30 minutes a day (that is, 30 minutes of actual productivity, not staring off into space and screwing around) and you keep at it, you can learn just about any language.

productivity: it's really not that hard

just do it.

nike,

I always just do it, even in the old-fashioned way with paper books at my desk. But there's no point in not trying to design better ways to get the job done, including the involvement of new technologies in the process.


Achab

Efficiency in language learning is key. I see so many people struggling with newspaper articles, looking up every tenth word, making lists and progressing so slowly. Efficiency creates intensity. With languages the more intense the learning the more likely you will make a real breakthrough.

Believe me, the voice of experience.

Define what efficiency means for you.

Efficiency means achieving the greatest possible amount of learning in the shortest period of time. The greater the efficiency of the learning process the greater the intensity of learning.

Trying to learn languages from uninteresting content is inefficient because the mind is turned off. Learning from interesting content is efficient because the mind is alert.

Learning a new language 3 times a week for one year is inefficient. Learning 6 times a week for 6 months is efficient.

Looking words up in a dictionary only to forget what you looked up is inefficient. Learning words in context is efficient. Learning phrases in context is efficient. Learning in an integrated manner is efficient.

Sitting in a classroom with 10-15 students is inefficient, albeit a necessary evil in many situations. Studying on your own is efficient. Listening, reading and learning words and phrases all from the same context is efficient. Writing using these same words is efficient.

Answering quiz questions, questions on grammar, questions on the meaning of what you just read or heard is all inefficient. It creates resistance in the mind of the learner. It disrupts the absorption of the language.

With efficiency the frequency with which you relearn what you have just learned increases. The efficiency feeds on itslef. A breakthrough becomes possible.

I agree, except for:

"Looking words up in a dictionary only to forget what you looked up is inefficient."

I have always relied on looking up words in a dictionary whenever I found an unknown term in a text, or in an online textual conversation, and sometimes even during the screening of a DVD.

My method is not trying to memorize the meaning of a word on the spot, but just reading what it means. Maybe the next day I will find that word again, and I will still not know its meaning. So I will look it up in a dictionary again. But after a number of times that I will have read the meaning of the word, finally I will also have learned it. This method have worked with me so far.

Plus, this is not learning words without a context, because I do not randomly memorize dictionary entries. I passively memorize the words that I find during my reading.


Achab

@ Achab:

>>FullRecall has a pattern of intervalls (1 - 8 - 20 - 38 - 110 - 365) that is different from the one used by Supermemo (1 - 8 - 20 - 34 - 87 - 210). I cannot get why FullRecall is more intelligent than Supermemo.

Another illustration which hopefully makes it more clear why FullRecall seems to be superior/more efficient than Supermemo (as mentioned earlier, figures might be different in reality):

The above indicated optimal repetition intervall by Supermemo might be ideal for "bad learners". In other words, if your memory's capability is much below the average learner's capability then the optimum repetition intervall might indeed be 1 - 8 - 20 -34 - 87 - 210.

However, if your memory capability is different to a "bad learner" your optimal repetition intervall will have to be different:

1 - 8 - 20 - 35 - 92 - 240 if you are an average learner
1 - 8 - 20 - 36 - 99 - 310 if you are an average learner
1 - 8 - 20 - 38 - 110 - 365 if you are a good learner

Supermemo doesn't take this into account - its repetition intervall is "dumb" and always fixed to 1 - 8 - 20 - 34 - 87 - 210 in our case.

FullRecall, however, is able to detect and "learn" which type of learner you are. As a result, FullRecall is able to adapt and optimise the repetition intervall if necessary and according to your abilities. So, if it finds out that you are a "bad learner" then the same repetition intervall as Supermemo's will be applied (1 - 8 - 20 -34 - 87 - 210). Otherwise, if you are an average learner, then FullRecall will change the intervall to 1 - 8 - 20 - 35 - 92 - 240 and so on.

In this respect, I find FullRecall far superior to Supermemo - especially if you want to learn big chunks of vocabulary an optimal repetition intervall makes a big difference in efficiency!

Using the dictionary is time consuming. Memory systems, including, or I would suggest, especially The Linguist, makes the learning of these new words , and phrases!, more efficient. Your whole language learning process, which is dependent on increasing vocabulary is sped up and made more effortless.

Okay DaVinci, I understand this second explanation of yours. Thank you. I will try to gather more information about the whole matter.

"SuperMemo doesn't take this into account - its repetition intervall is "dumb" and always fixed to 1 - 8 - 20 - 34 - 87 - 210 in our case. "

I believe this is incorrect. See:

The matrix of optimal factors OF used in Point 1 has been derived from the mathematical model of forgetting and from similar matrices built on data collected in years of repetitions in collections created by a number of users. ___Its initial setting corresponds with values found for a less-than-average student___. During repetitions, upon collecting more and more data about the student’s memory, the ___matrix is gradually modified____ to make it approach closely the actual student’s memory properties
http://www.supermemo.com/english/algsm8.htm

If you use a neural network, it has got to have some initial parameters, too. Otherwise, the NN will use totally "dumb" intervals at the beginning. It makes sense to start with parameters for a below-average user and then adjust them if the user is better than average.

FullRecall probably uses initial parameters designed for better-than-average users. The problem with this approach is that below-average users may be frustrated -- the intervals will be too long for them and they will keep forgetting stuff.

In any case, it is hard to optimize long intervals because of INTERFERENCE.

When the interval is 1 year, you are likely to encounter the word/phrase before the actual review date, which will disrupt the "perfect" schedule of reviews planned by your software.

So it doesn't really matter if the interval is 11 months or 13 months, as you will review the word/phrase sooner anyway.

@ Tom:

Thanks for sheding some more light. However, it seems as if there is some contradiction. Maybe somebody (the creator of FullRecell?) can clarify what is true and false.

A quote from FullRecall's website that I found at FullRecall website:

http://neodave.civ.pl/fullrecall/faq.php

Q: How does ANN in FullRecall work? Why (if at all) is better than a hard-coded algorithm found, for example in SuperMemo?
A: I don't know the exact details of SM algorithm(s) (I never was much interested in it), but important here is the main idea.
Algorithm in SM gets some data (e.g. number of repetitions, difficulty of item, current grade, etc. etc.) and returns next optimal interval it calculated. This algorithm, even if it's "smart" and corrects itself somehow, will be still dumb - it won't correct itself more than was designed for.

ANN on the other hand, is a black box from outside, was not designed like a normal algorithm, just learned.
ANN also gets some data (e.g. number of repetitions, grade, and so on), and returns next optimal interval it computed.
But on repetitions, FullRecall feeds ANN with more correct interval (i.e. if repetition was on time, and you hit "Poor" grade - last interval was definatelly too long, more correct one should be shorten). So FullRecall give ANN some feedback.
And ANN learns itself from this feedback, corrects itself, and tries its best to give more perfect intervals (perfect interval, in FullRecall terms, is when one hit "Good" grade on scheduled repetition, as the "Good" grade is considered the most optimal.)