This proposal suggests a new way to review LingQs using spaced repetition within the context of their original lessons. The system would recommend a daily set of lessons for review, chosen because they contain LingQs due for revision. Users could configure:
The number of lessons to review every day
The maximum lesson length to avoid overly long content
Courses to exclude from review
This would improve vocabulary retention by allowing users to review words and phrases in context.
I love this idea. Sometimes I remember the lesson and context where I read something, and sometimes not. It would be really good to have a way to automatically bring the lessons with many unlearned words up to the front.
Sound cool, a smart Lesson recommendation system based on SRS. There should be option to include/exclude completed lessons and include/exclude new lessons.
The big question, though, is how this is supposed to work. Should the recommendation be based on the amount of unknown words, the level of those words (1~4) or the amount of time passed since it has been read last? Or should it be a mixture of those, and how would it be weighted then? How is the lesson length supposed to affect this, if so? And should repetition be based with a preference of lessons with few unknown words for fast reading at high comprehension, so that a lot of words may get turned to known, or with lots of unknown words to feature lots of repetition? And how should the fact that there is no unified system to categorize the words as 1~4 or known be taken into account?
I donât want to play the devilâs advocate, but as much as the idea appeals me, too, I am not sure that a one-size-fits-all solution is going to work here.
Yea ultimately all that would be up to LingQ to decide based on their research, tech limitation, vision for their product and etc. It definitely wonât work for everyone so some options are needed but those can be added based on user feedback to get it close to working for most.
For me I will be interested to review/consume the lesson where the chance to move most yellow words to known is highest. So I want to review lessons (new/old) with as many LingQed(yellow) words that are ready to be reviewed (a certain X amount of time must has passed for both word and lesson similar to SRS). It would just be an intelligent guess since everyone is different. However the system could dynamically update recommendations based on how many words actually get upgraded or moved to known. Similar to ANKI FSRS system.
Actually if all that is too complex, a simple solution that could be easier to implement could be having more sorting/search features.
So add filter to sort by LingQ %. (yellow words)
Add way to filter out completed/new lessons.
Add filter to set new words %.
Add sort by date of completion/date of last reviewed/older than.
Sneak in an optional more compact layout for search result.
If possible add sort lessons by most LingQs ready for review (this might be complex I think) similar to OP idea.
With all that, user can sort all lesson by date of completion/review and by percentage of LingQ % (yellow words). User can then proceed to review lesson completed example⌠1 week ago with the most yellow words or LingQs ready for review
As for the other 2 options by OP, In the current search dialog we already have lesson length.
For a way to exclude course, user could tag them manually and this tag could be used to exclude by a better search system. Or there could be some other way to mark lesson/course for review and sort/filter accordingly.
This is not perfect, not exactly like how OP wants, there is no âsmart processingâ to recommend but I think it might be relatively much easier to implement, basically adding more search and sorting options and these could be added progressively.
In this way it is no longer a one-size-fits-all system, it is a you search for next lesson as you see fit.
If there is a way to save a few different search preset that could be cool too. OP could setup a search preset that they can click on every day to get a list of lessons according to the criteria set.
It also works just going over random lessons. You could also print out lessons and go over them page by page and return to old pages. A more random approach still works in the long run even if itâs not âoptimalâ
For a random non-SRS approach you can also hack it by adding the lessons you would like to revisit into a playlist then use the shuffle feature under playlist to let the system choose lessons to review randomly.
