Warning: this project has been deprecated
The algorithm itself has some inherent flaws, open this post to see why this lib is being deprecated.
If you are still looking for a space repetition lib, check memory-scheduler in lieu of this lib.
This is a JS implementation of a refined version of the SM2 space repetition learning algorithm invented by BlueRaja overcoming some of the inherent issues of the original version. Details about what these issuses are and how BlueRaja solved them can be found in this post.
$ npm install sm2-plus --save
;const DAY_IN_MINISECONDS = 24 * 60 * 60 * 1000;const getDaysSinceEpoch =Math;const TODAY = ;const testWord =word: 'test'update: TODAY - 17difficulty: 02interval: 100;console;
The output should be:
difficulty: 019interval: 1dueDate: TODAY+1update: TODAYword: 'test'
dueDate is the date the next time when the item should be reviewed.
It is good to be able to make a evaluation about how many times does the user has to study to totally remember a word in the best case.
Thus a simulation method was export to do this job.
The first argument is the initial difficulty and the second argument is the threshold below which a word can be seen as remembered.
The output would be like this, meaning that in the best condition, a user has to remember a word and select BEST 4 times to make sure that he has mastered the word whose initial difficulty is 0.3.
Each item should be stored as the following structure:
- difficulty: How difficult the item is, from [0.0, 1.0]. Defaults to 0.3 (if the software has no way of determining a better default for an item).
- interval: How many days should occur between review attempts for this item.
- update: The last date this item was reviewed.
When reviewing item, choose a performanceRating from [0.0, 1.0], with 1.0 being the best. Set a cutoff point for the answer being “correct” (default is 0.6). Then the item data can be updated using the way described below:
percentOverdue= (today - update) / interval) (2 <= percentOverdue)
difficulty = difficulty + percentOverdue * (8 - 9 * performanceRating) / 17 (0 <= difficulty <= 1)
difficultyWeight = 3 - 1.7 * difficulty
- 1 + (difficultyWeight - 1) * percentOverdue (for correct answer)
- 1 / difficultyWeight / difficultyWeight (for incorrect answer)