Flashcard Generator Concept Extractor

The card count here derives from concept extraction rather than page count alone. A page of dense technical prose - anatomy, pharmacology, statutory law - typically yields 7 to 12 separately testable propositions, while narrative humanities text yields 2 to 4. Each genuinely new concept expands into about 2.5 cards, which reflects the atomicity principle: one card should demand one retrieval, so a term usually needs a forward card, a context or reverse card, and often a card linking it to an adjacent idea. Concepts already known collapse to a single confirmation card, which is why prior knowledge cuts total workload faster than it cuts concept count. The scheduling half rests on the forgetting curve R = e^(-t/S), where R is recall probability, t is elapsed time and S is memory stability in the same units. Stability increases with each successful retrieval, which is why intervals expand roughly geometrically; 1, 3, 7, 14, 30 and 60 days is a standard ladder. Practical constraints bind harder than the arithmetic. Most learners cannot sustain more than 20 to 30 new cards per day without the review queue growing without bound, and reviews should absorb roughly 80 percent of session time. The projection assumes well-formed cards reviewed on schedule; cards packing several facts onto one side, and skipped review days that let items fall off the curve, both push real retention well below the modeled figure.

Hermann Ebbinghaus, working in Berlin, memorized lists of nonsense syllables on himself and published Uber das Gedachtnis in 1885, producing the first quantitative forgetting curve and the first clear evidence of the spacing effect, that distributed repetitions outperform massed ones. His findings sat largely unexploited for decades. Sebastian Leitner, a German science journalist, made them operational in his 1972 book So lernt man lernen, describing a physical box system in which a card answered correctly advances to a compartment reviewed less often while a card missed returns to the first box. Computerized scheduling began with Piotr Wozniak, who worked out the SM-0 paper algorithm as a student in Poland in 1985 and released SuperMemo for DOS in 1987 carrying the SM-2 algorithm. Because Wozniak published SM-2 openly, it became the de facto standard: Damien Elmes built Anki on it in 2006, and Mnemosyne and many successors followed. The experimental backing arrived in parallel through Robert Bjork's desirable difficulties framework and the retrieval practice experiments of Henry Roediger and Jeffrey Karpicke published in 2006, which showed testing itself, not merely restudy, drives durable retention. Contemporary schedulers such as the free spaced repetition scheduler fit stability and difficulty parameters to an individual review log rather than applying a fixed formula.