As K1 starts to take shape and the new UI goes live I am strongly encouraged to keep working on this and get it done. Here is a short summary of what we have new in K1 so far.
- New responsive UI is being tested, so far it looks good on both desktop and mobile.
- Dictionary cards have an ent_seq (entry sequence) field which is intended to be used for sorting search results.
- Dictionary frequency field is auto-calculated based on ftags and dictionary language using a new tag processing library.
- Dictionary search results can be added as Study Cards via an “Add to Deck” button.
- Study cards have a status field added; set to ‘new’ for new cards, for example.
- When editing cards the language appears below “Keyword”, etc.
- Study Cards use a new algorithm foundation based on SM-2.
The big story here isn’t the move to SM-2, but more on that in a moment. The big thing here is you can click to add flashcards based on dictionary entries. The system remembers that it is based on a dictionary entry forever, and will update that flashcard if and when the dictionary entry changes.
But if you don’t want to add cards that way, you don’t have to! You can always enter your own cards if you find it helps you remember.
SM-2 aka “Time Bubbles”
We’ve finally switched away from the fscore system and into true spaced repetition. The algorithm is basically SM-2 (what Anki uses) but without the overly complex configuration options that Anki has. The fact is, very few people make good use of things like “starting ease” or the ability to set graded intervals by hand. These are things that the program should be able to figure out about you anyways, if it’s smart and well-written. So we will start with similar defaults and allow the program to learn how you learn. You will teach the program how to rate you. For example, if seeding a card as “known” tends to indicate you will rate it as “known” a second time, on average, then the program will learn to increase the interval for “known” the first time you enter it. Similarly, if telling the system you forgot a card tends to mean the next time you see that card you will mark it forgotten again, then the program will learn to show you that card more and more often until you either suspend manually or start getting it right.
Our next goal is to move the algorithm closer to a SM-6 and/or SM-8 “matrix” style, or to come up with our own unique system (the SM way is by far not the most optimal way to do things, it’s just the best way to deal with memory curves). Coming up with our own unique system may in fact just be using SM-8 and weighting the results from various areas of the site (like automated cloze tests, for example) feeding back into the flashcard algorithm. There are also some interesting features in SM-11 we will be adding, like giving a proper consideration to early review intervals and ‘late’ intervals. For example? If you get it right after a very late interval, the program will keep that in mind when setting your next interval. But if you get an early review wrong, the system won’t move your interval forward beyond what you set it, because you haven’t yet demonstrated you can remember the card that long. Hey, you chose to do an early review, it’s your own inefficiency, we are of the opinion that you can do what you want, even if it’s inefficient, it’s your life! And sometimes people just like studying a little inefficiently, it can be fun and add to your confidence.
Lots of exciting stuff coming soon! Don’t touch that dial!