I assume that any spaced-repetition alogorithm must always
"underestimate" the interval for the next presentation. If the
algorithm overestimates, then the user forgets the card. Three years
work gone. Start again :)

Due to the need for underestimation, I think the cards might take
longer than necessary to push out into the future. Here's a suggestion
for possible improvement.

The artillery man always kills you within three shots, because he
deliberately "straddles" you. Here are the rules of the artillery
man's straddle:

1)  He will always try to hit you with his first shell. If he does,
it's an unexpected bonus.

2)  If he misses, he knows by how much he missed you.

3)  He will NOT try to hit you with his second shell. (This could
result in several misses.)

4)  If his first shell lands in front of you, he will deliberately
place his second shell behind you. He now knows the amount of error in
his first and second shots.

5) Ciao. RIP :(

So, my suggestion is to allow the user to insert some "straddle" cards
in a deck. These cards can either be cards that the user genuinely
wants to memorize, but is prepared to sacrifice. Or, they could be
cards of equivalent difficulty to the "real" cards, but the user does
not care if the card is forgotten. Approximately 10% of cards could be
straddle cards.

The user would need to identify each straddle card. Each time
Mnemosyne encounters such a card. then the algorithm is allowed to
calculate a straddle estimate on that card. At the next presentation
of that card, the user will either remember or forget.

If the user remembers the straddle card, then a straddle did not
occur. (In fact, this is a wasted effort.)

If the user forgets, then a straddle has occurred. This information
can now be used to alter all the estimates for all the non-straddle
cards. (They must still underestimate the interval, but hopefully by
less than predicted by the usual SM2 algorithm.)

Why would we need 10% of cards? To provide greater accuracy in the
straddle estimate. Maybe an exponentially smoothed average of the
straddle intervals could be maintained.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"mnemosyne-proj-users" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/mnemosyne-proj-users?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to