Retention rate and question about cloze deletion

149 views
Skip to first unread message

J Clark

unread,
Apr 13, 2012, 1:06:41 AM4/13/12
to mnemosyne-...@googlegroups.com
Hello all,

Firstly I'd like to ask: is it possible to change the intended retention rate of 90% that I believe Mnemosyne is set to by default?  It would be excellent to have different rates for different categories (or even cards?), especially.  There are things which are more or less important in my database (one database which contains all of my categories, as Mnemosyne recommends) and I would like to have the algorithm deal with this difference in priority rather than artificially "not worrying about" some cards and performing unscheduled reviews on others.  I suppose that's optimal, bu if I could just temporarily bump up the rate for the whole database that would be good too.

Secondly, (and this might be a dumb question), I have a vague memory of reading about some form of automated cloze deletion, does that exist?  Right now I just copypaste the sentence in and replace a word with ___, sometimes more than once.

Okay, actually I have another question about cloze-deletion.  Is it known in some quantifiable way if cards made from the same sentence or image appreciably interfere with each other by "giving away" the answer to the other cards in that family?  I've often found that since I make them at the same time they repeatedly come up together and I've wondered if I should avoid this problem by editing such cards to give minimal context and therefore not "spoil" the others, but I feel this itself would also be bad for memory.  Any opinions welcome!

Thanks, and thanks Peter for a great program.

Jeff

Peter Bienstman

unread,
Apr 13, 2012, 4:50:04 AM4/13/12
to mnemosyne-...@googlegroups.com
On Thursday, April 12, 2012 10:06:41 PM J Clark wrote:
> Hello all,
>
> Firstly I'd like to ask: is it possible to change the intended retention
> rate of 90% that I believe Mnemosyne is set to by default? It would be
> excellent to have different rates for different categories (or even
> cards?), especially. There are things which are more or less important in
> my database (one database which contains all of my categories, as Mnemosyne
> recommends) and I would like to have the algorithm deal with this
> difference in priority rather than artificially "not worrying about" some
> cards and performing unscheduled reviews on others. I suppose that's
> optimal, bu if I could just temporarily bump up the rate for the whole
> database that would be good too.

In 2.0, you could write a plugin to achieve this. However, I'm not sure it's a
good idea to have this functionality in the core program (i.e. not in an
optional plugin), as tweaking the numerical parameters of the algorithm could
have unexpected consequences over time.



> Secondly, (and this might be a dumb question), I have a vague memory of
> reading about some form of automated cloze deletion, does that exist?
> Right now I just copypaste the sentence in and replace a word with ___,
> sometimes more than once.

It's the cloze card type in 2.0.

> Okay, actually I have another question about cloze-deletion. Is it known
> in some quantifiable way if cards made from the same sentence or image
> appreciably interfere with each other by "giving away" the answer to the
> other cards in that family? I've often found that since I make them at the

> same time they repeatedly come up together.

If you use the cloze card type in 2.0, Mnemosyne knows that they are derived
from the same information and not schedule them on the same day.

> Thanks, and thanks Peter for a great program.

You're welcome!

Peter

J Clark

unread,
Apr 13, 2012, 8:13:49 PM4/13/12
to mnemosyne-...@googlegroups.com
Huh.  It seems that I should upgrade to 2.0 ;)
Is there a stable release of that which I can download somewhere?  Mnemosyne-proj only mentions current version 1.2.2.


>> Firstly I'd like to ask: is it possible to change the intended retention
>> rate of 90% that I believe Mnemosyne is set to by default?  It would be
>> excellent to have different rates for different categories (or even
>> cards?), especially.  There are things which are more or less important in
>> my database (one database which contains all of my categories, as Mnemosyne
>> recommends) and I would like to have the algorithm deal with this
>> difference in priority rather than artificially "not worrying about" some
>> cards and performing unscheduled reviews on others.  I suppose that's
>> optimal, bu if I could just temporarily bump up the rate for the whole
>> database that would be good too.
>
>In 2.0, you could write a plugin to achieve this. However, I'm not sure it's a
>good idea to have this functionality in the core program (i.e. not in an
>optional plugin), as tweaking the numerical parameters of the algorithm could
>have unexpected consequences over time.

Well, I doubt that I'll be writing my own plugins anytime soon; I'm not a "power-user" or programmer by any means.  And (keeping that in mind), I don't really know what you mean by unexpected consequences.  I thought that in Supermemo even as early as SM2 it was possible to modify your target rate of memorization (correct me if I'm wrong about that, I didn't really spend much time there).  I kinda thought that it wouldn't really be a difficult thing, just sort of "scaling up the graph" representing the intensity/frequency of reviews.

I thought of this because I have tests coming up.  It's not a major issue, and Mnemosyne is certainly not my only method of review, but I'm interested in hearing what I might have overlooked here.  If it's simply a programming hurdle then there's not much I can say, but from what you've said it doesn't sound like that's the case.

Jeff

Peter Bienstman

unread,
Apr 14, 2012, 2:33:38 PM4/14/12
to mnemosyne-...@googlegroups.com
On Friday 13 Apr 2012 17:13:49 J Clark wrote:
> Huh. It seems that I should upgrade to 2.0 ;)
> Is there a stable release of that which I can download somewhere?
> Mnemosyne-proj only mentions current version 1.2.2.

So far, it's only announced on the mailing list here.

You can find the installer here:

http://users.ugent.be/~pbienst/pub/mnemosyne-beta-11b-setup.exe



> Well, I doubt that I'll be writing my own plugins anytime soon; I'm not a
> "power-user" or programmer by any means. And (keeping that in mind), I
> don't really know what you mean by unexpected consequences. I thought that
> in Supermemo even as early as SM2 it was possible to modify your target
> rate of memorization

Not really sure that's a valid argument: Supermemo is extremely bloated and
full of functions which I find confusing for new users :-)

> If it's simply a programming
> hurdle then there's not much I can say, but from what you've said it
> doesn't sound like that's the case.

It's not that I'm against implementing anything like this, but I believe it
should be in a optional plugin, and I'm afraid my Mnemosyne todo list is
rather full at the moment.

Here's an alternative workflow I suggest you try out once you move to 2.0:

2.0 supports saving a selection of your cards (e.g. a combination of certain
tags) to what's called a 'saved set'. Make one of these for your important
cards, and one of them for your non-important cards. Study your important set
each day, and only study your non-important set when you feel like it.

Cheers,

Peter

J Clark

unread,
Apr 15, 2012, 1:21:50 AM4/15/12
to mnemosyne-...@googlegroups.com
>> And (keeping that in mind), I
>> don't really know what you mean by unexpected consequences.  I thought that
>> in Supermemo even as early as SM2 it was possible to modify your target
>> rate of memorization

>Not really sure that's a valid argument: Supermemo is extremely bloated and
>full of functions which I find confusing for new users :-)

There was a reason I didn't stay with it for long!  I was going to include a quick jab at SM in that post (something about how it required its own set of memory cards to use it; actually now I think I remember that it literally has that set) but I figured it'd be unnecessary.

That said, I was referring to the fact that the algorithm in SM2 (which I think is what Mnemosyne uses, at least as a base) could be modified in this way, but nonetheless I take your point and I'll shut up about it.

I look forward to playing around with 2.0, it sounds like a big update.  Are we expecting an official release soon?

Thanks for your patience ;)

Jeff

Peter Bienstman

unread,
Apr 15, 2012, 4:52:54 AM4/15/12
to mnemosyne-...@googlegroups.com
On Saturday 14 Apr 2012 22:21:50 J Clark wrote:

> I look forward to playing around with 2.0, it sounds like a big update.
> Are we expecting an official release soon?

It's not yet officially released, but don't let that stop you, as it's very
stable and only lacks some minor polish.

Peter


Reply all
Reply to author
Forward
0 new messages