This website uses cookies. Find out more or .

Questions and Answers

Question

0
points
Asked 157 months ago by ribenguniang  valid Edit

Hi,
If I look at your Q&A entries, it seems there still are some points that are not clear to everybody about your startegy to determine the next repetition time/date of a word. I also have a question concerning it:

Is it so that a word considered as 'known' by your application will never be asked again?

Add comment  
This question has been viewed 29 times.
Answers
1
point


Hi Klaus,

Thank you for your very precise explanations. That is about what I had understood, ... and actually what I feared. Let me explain.

You see, my mother language is French, I spoke (nearly) only French during 20 years (except college, of course). Then I moved to a German speaking city, so I speak German at work and in addition Swedish at home. When my daughter was born a few years later, I often wanted to tell her the French word for this and that, so that she could learn. I was frighten to see, that I often did not (easily) remember it, especially the name of things I one needs a bit more rarely.

So I know by experience that a word is rarely permanenttly memorized, especially if there is nothing in the environement you are living in, that reminds you of those somewhat complex Chinese words and their graphic representation.

Couldn't it be an option, to give the user the possibility to include a few of those 'memorized words' each time he repeats his vocabulary. One could imagine for instance some settings where the user can say; include 1% of the 'memorized words' into the list of words to repeat, but at most 15 and at least 5. Something like that....
As I wrote in one of my earlier comment, I know around 2000-3000 hanzi, but frankly said, if somebody would ask me right now to write '必须' or '需要', quite basic words actually, I think I would fail...
I think implementing this would be very easy, one would only need a new flag associated to the 'memorized words', to tell if a word has been already reviewed or not in this 'round'. Next round, you just  inverse the polarity. Of course if you fail, the word is automatically re-activated, or the user could be asked to confirm if he wants to re-activate it (maybe re-practicing a forgotten 'memorized words' once in all modes could also be sufficient..)

考虑,考虑。。

Answered 157 months ago by ribenguniang valid Edit
Add comment
0
points
I worry the discussions are not easy to follow as it may have different aspects. Below I will try to summarize the methodology, which is automatically applied by the trainchinese system.

Each time a card is presented in one of the training modes, you decide if you \'know\' or \'are not sure\'. Only after passing the conditions of enough \'know\' against the \'not sure\' which is required for all of the enabled modes of a card, then the card will be considered learned once.
After being learned the card will sleep, and then appear again after a certain sleeping time.
A card will need to be learned in above way several times before it is considered \"fully learned\" or \"memorized\".

There are two important parameters which are configurable by the users in their settings:
  • You can define how many times you want a card to be shown before it is not shown any more because it is considered memorized.
  • You can define the base time period how large the first sleeping time should be (you may prefer to have short intervals, or longer intervals), the actual sleeping time will be calculated on a card by card basis using that parameter and the information we have on how you have been learning that card.
We are basing the decision on when to show words again on those two parameters. The sleeping time will be increased for each of the following times a card sleeps.

The setting of the basic sleeping time and the number of repetitions may depend on the personal ability to memorize Chinese, which may depend on many aspects. That is the reason we let the user configure that. People with bad memorization skills will need short sleeping times an many repetitions, and those good ad memorizing can select longer sleeping times and fewer repetitions.

Once a card has slept the number of times you defined (in the settings), that card will be considered \"Fully learned\" or \"Memorized\". Such cards continue to be in your vocabulary lists, but will not appear as flashcards - unless you decide that they should be \"learned again\". Which can alwasy be selected on a card by card basis pressing one button.

One note on the number of times a card sleeps: Users have the option to Wake Up sleeping cards before their due date, in that case the corresponding sleeping period will not be considered for reaching Fully memorized status, as we assume that the suer feels insecure about that card.

Just as a reminder: a card will be trained according to the different Chinese training modes (that could include Reading, Oral recognition, translation and writing). Each card can have diffrent settings for this modes.
Answered 157 months ago by klaus valid Edit
Add comment
0
points
I (somewhat) second this proposition. I also just recently experienced the same as ribenguniang, that you never can really say \"I\'ve learned this word now and won\'t ever forget it\". Of course in the personal settings you can change it to a currently maximum of 6 repetitions of the words before they\'re considered \'learned\', with a starting interval of 5 days that would mean an interval of 160 days at last which makes it possible to sparsely add a few old words into your learning stream. But then there\'s the psychological thing with \"I want to tick these words off as learned\"which will take forever until one word gets there (5+10+20+40+80+160 days = nearly a year). I personally need this confirmation/feeling of achievement to make progress.
In other systems of spaced repetition I know, once you don\'t know a word that\'s been presented to you, the learning progress gets set back to 0%, regardless how close you were to the \'nearly learned\' status.
I don\'t know a perfect solution to this. Maybe change the progress statistics to show how many cards you have in each percentage level (according to your wording there), maybe other people can think of something. This is more something to get the discussion started/going...
Answered 157 months ago by heiks valid Edit
Add comment
0
points


Hi also support my own opinion ;-)

Actually I would combine my proposal with klaus\' proposal. Each word (inclusive memorized words) is associated with some information like (I just guess) the number of times a user repeated that precise word (in each mode) and how many times he failed. So I think, the app can figure out how difficult it was for that user to learn a given word, and how probable it is that he will forget it again, even though it is marked as memorized. And so the app can determine if a given memorized word should be included in the bunch of words to be repeated.

I also agree with heiks, if the user fails to recognize a memorized word, the cards shall not come back to 0%. Actually I would put it to the highest level, the level a cards reaches just before it goes to \'memorized\'. From that moment on it is processed normally again.

I think it would be worth a try. The trainchinese team surely has a lot of statistics about the user\'s habits and leaning progress, how often memorized words are re-actived, aso. Maybe they could, in the light of this discussion, to try find out how such a feature (or similar) could be implemented.

Regards

Answered 157 months ago by ribenguniang valid Edit
Add comment
valid
   Edit
klaus lautet:
Are thinking about it ...
0
points

I\'v been thinking a little bit about this \'feature\'.
Last month I was in Beijing to lear Chinese three weeks, and I took with me around 600 paper Flash Cards (words hard to remember). I practiced a lot, and was actually happy about my progress.
This week I looked at those words again.... Deception: 20% were already gone ...
I am convinced, there MUST be a way to integrate the repetition of \'memorized words\' in your process.
I am trying to read Chinese as often as possible, listen to audio books or so. But fact is; we don\'t have a Chinese environement were we can meet \'memorized words\' in our daily life. For most of us, our Chinese environement is our Flash Cards, and they shall also reminds us of memorized words!

I tought of a very simply solution:
Imagine that the \'memorized words\' is not a list but a folder.
Within this folder the system would create three lists:

- easy
- normal
- hard

The user cannot moved words from one list to the other within the \'memorized words\' folder.
Words can only be \'degraded\', i.e come back to a \'not learned\' state.

As I said this list are automatically managed by the system, evaluated based on words statistics, either as a \'background\' process, or each tme the user makes an \'update cards\'.

The system automatically assign to those lists a fix percentage of known words, and a fix time for the completion of one repetition cycle.

level      ¦    %-age  ¦ cycle duration
-----------+-------------+-------------------------------
easy      ¦      1/2      ¦  500 days
normal   ¦      1/3     ¦ 300 days
hard      ¦      1/6      ¦ 150 days

Looking at my personal paper FlashCard situation, I have around 4000 words.
So each day I will have the following amount of \'memorized\' words in addition to the normal processing.

easy     : 2000 / 500 = 4 words
normal : 1330  / 300  = 4 words
hard     : 666    / 150 =  4 words

So I will have 12 additional words everyday for this \'background\' work, it is really not excessive, I usually have 70-100 words each day to repeat.

What do you think of that solutions?

The user would still have the \'good feeling\' of seeing his progress.
He has nothing to configure for this automized long term repetition.
In addition, there is no additional word related values or flags to store in your DB (no upgrade necessary).
The only thing we need are three dynamically managed lists, managed by the system.
In my opinion, this solution would combine all advantages of what we discuss in this thread.
I think this is also quite easy to implement, so it would be worth thinking about,

Of course this proposal should be combinable with an other suggestion I posted: \"Moving form one level to the other\" (at least for the words marked as \'hard\').

Kind regards
Always at you disposal for new ideas ;-)

Answered 156 months ago by ribenguniang valid Edit
Add comment
0
points


Hi there,
My last comment on this feature remain uncommented. I really think this feature is absolutely necessary. I absolutely have to repeat my \'memorized words\' of my paper Flash Cards otherwise I will forget everything quite fast. So I expect this feature with my electronic Flash Cards too.

Answered 155 months ago by ribenguniang valid Edit
Add comment
valid
   Edit
klaus lautet:
We agree that it would be helpful if we provide an easy way to select words from the "Memorized Words" based on the likeness to have been forgotten. As many users do not like the idea of "automatic" addition of words to be learned, we think any addition of words "likely to be forgotten" needs to be initiated by the user.

What we would think of is to have a way to sort all of the Memorized words based on how long a card was "memorized" already and how difficult it was to reach the "memorized" status. Once the list of memorized words has been sorted by 'likeliness to have been forgotten', then a user could just select how many of them he would like to 'review'.

We are considering a differentiation between 'Review' and 'Start learning again'. A 'review' would move the card into the highest level of training, that is if during the review all modes are passed correctly the card immediately will return to 'memorized'. That compares with 'Start learning again' where cards will move to the initial stage of learning and will need to go trough all the sleeping cycles before they are 'memorized'. Obviously if the user had difficulties during the review the card will sleep and will be then trained again (and again..). The review option would only be available for cards that have been memorized.
valid
   Edit
ribenguniang lautet:

Again your solution is very pragmatic, using already available mecheanisms.

First a comment: I do not quite agree with your statement that users do not like the automatic addition of words. They don't like to have random words automatically added, without any context and without respect to their actual situations, like the book they are reading, deng deng ...

Now let's look at your solution. I am just wondering how it woud work in reality, so I take my personal case as an  example.

If I had started my Chinese study with your app, I would now have around 3000 words as 'memorized' and 1500 words beeing processed (just an assumption).
I suppose I would like to have at least 2000 words repeated out of the 'memorized' words.
Of course I cannot have all those words suddenly added simultaneously into the list of words to be processed, this has to happen progressively.
So how is this done?  What are the selection criteria (weighted randomness?) . How many words will be added dayly into the 'highest level'? Can the user configure it? 

Of course I also would like to have a greater probability (but just a probability) to be asked the less well known words (out of the list of words to be reviwed)
This is what I wanted to take into account in my proposal, which aimed at repeating all memorized words, respecting their 'degree of memorization'

So maybe it would be  good if you could refine your solution a bit, in order to make it work a bit more in the direction of my proposal. Maybe for instance one could select the words one do NOT want to repeat, and then apply some weighted probability mechanism for the daily addition into the list of words to be repeated.

Note that defining the words that you do NOT want to be repeated is in my opinion better than the contrary for two reasons:
1. It is probably the shorter list
2. New words just 'upgraded' into the 'memorized' state, will be automatically added to the 'review' list, without any user configuration.

I would be glad if you could consider my proposal, this is an important feature.
I think it is worth evaluating different options.
I would really appreciate if other users would take part in this discussion too. I am sure that together we can find a pragmatic, efficient user friendly solution!

0
points
I've been thinking a little bit about repeating known words. I think that even the new advanced learning mode will soon be available, it is still necessary to repeat the words we know, but it shall be quite a quick repeat, nearly like a hand free mode.
I hope that you are already working at the solution you mentionned above with the sort possibility. The the user can simply say: ok I want to learn the 'top most' 1000.
But than I wonder how those words are integratred into the 'daily routine'. I would not mix up the normal learning process and the repeating process. As I said before, the repeat process should be quite fast.
So I would say your app shoul have three mode of operation:
- a training mode: the mode we have now
- a handfree mode
- a repeat mode.

In the hand free mode, the selected word list(s) are just read out loud (English->Chinese, Chinese-> English, according to user configuration). A pause button shall be available. The words to be reviewed in hand free mode are any list(s) of words inclusive the 'sorted list of memorized words'. In hand free mode the word statistics are not changed. The app shall remember where I stopped and continue from that point the next time I use it.

The repeat mode is nearly like the hand free mode, except that
- statistics are changed and a word even can be 'degraded'
- there are two : 'pause' and  ' I forgot'

As in the hand free mode, it shall remember wehre I stopped the last time I repeated.
In this way you can quickly go through all your vocab, just saying 'I forgot' now and than.
As an addtional feature I would say that in this mode words should be repeated in three times:
- now
- in 10 days ask again
- in 30 days ask again

I would be please to be informed about your progress with this feature and I would be very happy to be your tester if you need a feedback during the implementation.
Answered 155 months ago by ribenguniang valid Edit
Add comment
valid
   Edit
klaus lautet:
Thank you for the detailed suggestion - we will start working on the "Review learned" functions once we have finished our homework regarding the improved algorithm and statistics functions.
valid
   Edit
klaus lautet:
We have finished to implement the following functions for the re-learning or review of memorized cards:

1) Instead of Learning again a card (which would add a card to be learned trough the whole learning cycle) now it is possible to "Review" a memorized card. That will effectively put the card into the last of the cycles, so if "learned" then it will be again in the list of memorized cards.
If the user is using the advanced algorithm of calculating sleep times (in Settings of this page), and the user made errors when Reviewing the learned card, then the card will not go to Memorized" but will need to be reviewed again in the problematic modes before it can become Memorized again.

2) We introduced a function that will order the memorized words according to a combination of two criteria:
One is the general difficulty the user had to memorize that word
the other is how long ago the word was memorized.
valid
   Edit
klaus lautet:
The order of the cards considered to be possible for review is calculated based on a series of test accounts we are using and may be necessary to be adjusted. So any comments on that are welcome.

On the other hand we do not plan to add any automatic functions that would add memorized cards to the users active cards. Users who like to add memorized cards will need to go to the website and add them form there (given the functions now it is neither difficult nor time consuming).
0
points

Answered 151 months ago by Tranimal blocked Edit
Add comment