[lojban] Logflash for UNIX

[Taral <taral@taral.net>]

Okay, I have the modules needed for Logflash on Python. Attached is my
program plan, along with a few algorithmic ideas. Comments anyone?

Taral

------------------------------------------------------------------------
eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/3864/3/_/17627/_/957900439/
------------------------------------------------------------------------

To unsubscribe, send mail to lojban-unsubscribe@onelist.com

init:
    load data files
        wordlist
    login screen
    if user exists
        load user files
            mode
            piles (indexes? compression?)
            word test number
        goto lesson overview
    else
        create user files
        set new words mode
        set word test number to 0
        goto main menu

main menu:
    (menu)
    goto lesson
    show wordlist
    lookup word
    goto options menu
    change mode
    restart all words
    exit

options menu:
    (menu)
    max lesson size
    error practice factor
    recog typos
    recall typos
    dropback pile
    new words active
    goto main menu

lesson overview:
    display piles (arrow on first pile of piles_to_do) (special UC!)
    (menu)
    goto do_lesson
    goto session select menu
    toggle new words active
    goto main menu

do_lesson:
    until user wants to quit:
        while piles_to_do:
            get pile from piles_to_do
            if pile empty:
                continue
            if pile size > max lesson size:
                put (max-size) least important words back in original pile (*)
            test words in pile (*) (!)
            put correct words in next pile
            put errors in error pile
        if mode complete:
            set next mode
            goto lesson overview
        set piles_to_do to all_piles (according to mode)
        rotate UC piles if applicable (not needed if age tracking)
    goto lesson overview

(*) Uses frequency data, error rate, and age for prioritization.
    Age is time/number of words presented since word last exited UC/NW pile
(!) Error words are tested many times until correct. Number of times based on
    error practice factor, frequency data, and error rate (w/ time-decay) if
    available.
    Under Control pile has high randomization factor, others low.
    (variance/range based?)

Ideas:
Typo checking based on all possible words, not just correct answer
Confusion matrix (sparse)
Accumulated confusion data for general statistics
Confusion clearup mode (multiple choice)