Re: memory management in orgmode (supermemo)

["Alexander Laertes" <ablaert@gmail.com>]

Hello Pere,

On Mon, 21 Sep 2009 10:36:12 -0400, Pere Quintana Seguí wrote:

> The other thing that supermemo implements is incremental reading
> http://en.wikipedia.org/wiki/Incremental_reading
>
> I don't know much about it, as apparently only supermemo implemented it.
> But I guess it would be also well adapted to org-mode.
>
> Did you ever try it? Do you think it would be a nice addition to  
> org-mode?

As the Wikipedia article explains, incremental reading (IR) is a mechanism
that eases the input of new material that is to be processed at various
stages, incrementally.  First, you read an article's content normally;
when you think you've come across an interesting portion of text (say, a
sentence, paragraph, or section), you extract it (think highlighting);
these extracts will be new sub-documents (in SM parlance, topics) that
will be shown at a later date, as scheduled by the spaced repetition
algorithm.  When these topics come to your screen, you will re-process
them (e.g. by shortening and rephrasing of sentences) to finally convert
them into typical question/answer items that are common to most SR
products (SM, anki, mnemosyne, etc).  You can think of it as an integrated
pre-processing step to build Q/A material.  Its incremental nature helps
tackle difficult material, as it is possible to postpone topics that
require supporting data or better explanations before coming back to them,
as well as counter factors such as frustration and boredom.  On the other
hand, the fact it is fed to a scheduling process (spaced repetition) is
quite beneficial as you know it will come back to you, and because each
exposure to the material aids in establishing the memory trace, making
future recall easier.  For this reason, incremental reading techniques
seem to benefit the most from a spaced repetition algorithm (a la
Supermemo).

For org-mode to support IR techniques in a rudimentary way, a notion of
document and sub-documents/extracts seems vital; if these could be
separate sections in an org-file (corresponding to a single article), then
to support the flow of information in IR there could be a mechanism to
make copies of the highlighted material (say, the active region) into a
new subsection:

==== file: ATitle.org ====
* Article: A Title    :article:

Lorem ipsum [mark]dolor sit amet[point], consectetur adipisicing elit
====

M-x org-make-extract

==== file: ATitle.org ====
* Article: A Title...    :article:
** A Title: Lorem ipsum  :topic:

dolor [mark]sit[point] amet
====

M-x org-make-cloze-deletion
M-x org-dismiss-section

==== file: ATitle.org ====
* DISMISSED Article: A Title...    :article:
** DISMISSED A Title: Lorem ipsum...  :topic:
*** A Title: dolor        :item:
**** QUESTION
dolor [...] amet

**** ANSWER
sit
====

M-x org-drill

====
** A Title: dolor     :item:
*** QUESTION
dolor [...] amet

*** ANSWER...
====


More ideas:

Topics with memorized/dismissed states could be worked around TODO states
or tags, which should not propagate to children (memorized means "in the
learning process").  Categories (as in a knowledge tree) could be
implemented with #+CATEGORY or tags.  Encoding of parameters in items
(such as next repetition date, repetition history, A-factor, forgetting
index, etc.) could be worked around drawers (though some of this data
perhaps shouldn't be user-editable).

An old version of the SM algorithm (SM-2, current is SM-11):
http://www.supermemo.com/english/ol/sm2.htm  (Incidentally, anki and
mnemosyne are based on SM-2, therefore some source code should be
available)

Addendum: to experience incremental reading in Supermemo, keep in mind the
only IR-enabled versions are Supermemo 2000 or later for Windows.  There
are trial versions available for download.  Also, note that the knowledge
management abilities related to IR take a bit of time and focused practice
to develop.

Alex