I'm somewhat confused about the way Skritter handles the distinction between learned characters and learned words. Sometimes they appear to be treated separately (as in [1]), sometimes they are not even recognized if previously learned in combination (as in [2]).


[1.] Knowing how to write a character, but not knowing it in context.

Chengyus are a good example: When I'm studying a Chengyu like 莫名其妙, the characters themselves aren't the difficulty. Still, I may confuse 莫 with 陌. Skritter will regard 莫 as unknown and then later ask me to learn it as a single character, although I have been consistently able to write it.

I still constantly need to review easy characters, because I confuse them with other characters if used in combination. Lately I've sometimes adopted a method of manually marking the character as "known" and the word as "unknown".

Most frequently, it happens with "jie":
解 (spent 32 min), 结 (spent 25 min), 接 (spent 25 min)


[2.] Knowing a character as part of a word, but still "unlearned" as a single character.

After having passed the 2500 characters mark, I'd really like to "fill the gaps" by simply running through some of the nice character lists (http://www.skritter.com/vocab/tags?tag=characters).
However, many of the characters I definitely know are still shown as completely unlearned. Examples are 应酬的酬, 儒家的儒，亭子的亭 and so on.
Why aren't they recognized this time? Aren't characters treated seperately? I thought these characters already have their own database entry since I've already been studying them. The Viewer (http://www.skritter.com/vocab/viewer) and my experience seem to suggest each character and each word gets its own entry, still these characters aren't recognized as being learned.


Thanks in advance for your help!

"I still constantly need to review easy characters, because I confuse them with other characters if used in combination. Lately I've sometimes adopted a method of manually marking the character as "known" and the word as "unknown". "

I've been doing this for a good while now. It's rather become second nature.

I think Nick has some logic in there to avoid really common characters being marked unknown when it's clear from context that it's selection of the character that's the problem, not how to write the character itself. On a related note, when I had "一" come up with tone study after I got some tone changes wrong too many times Nick added some additional logic. The problem wasn't that I didn't know the tone for yi but that I didn't know the next tone (so couldn't apply the tone change rule to get the correct result).

Unfortunately it's not always clear when it's the choice of character or the character itself. I do the same thing (manually mark words wrong and the characters right). Of course, if you get the character right in isolation (say) 5 times in a row the system should have enough data to make the choice itself (once all the logic behind it is in place) and reviewing a character 5 extra times is likely to be about equivalent to the extra time it takes to manually grade (and more useful as it is reviewing characters instead of pressing buttons).

Nick, am I misquoting and/or being wildly optimistic about the intelligence of the marking algorithms?

"reviewing a character 5 extra times is likely to be about equivalent to the extra time it takes to manually grade (and more useful as it is reviewing characters instead of pressing buttons)."

就是！So something like a simple additional keyboard shortcut could be helpful.


What about #2 ("Knowing a character as part of a word, but still "unlearned" as a single character")?
Do I need to start from scratch for all those characters that should already be included in the list of characters learned?

I embarrassingly have not put in the logic for writing and tone prompts to automatically differentiate like that yet. It's been on my list since forever ago. But yeah, it will work like that once I do it, so it will eliminate the problem of grading the character and word differently (which is a decent workaround for now, but perhaps slower than not doing it--anyway, I do it).

I do have it for reading prompts (and the tone prompts that they contain), so at least that works the best way one would expect.

The characters in those lists haven't been added by themselves, so they won't be prompted individually, but progress is being tracked. If you added them, they would activate solo and you'd be able to see that they are learned. You can have this done automatically if you want, with the "Also Add Characters When Adding Words" checkbox on the language settings page. (It doesn't apply retroactively, though.)