Comment by aaldrick
1 day ago
Not sure I understood the gist, but let me try summarise what you do generally (i.e. not just for language learning).
- Instead of flashcards, you read content/write notes. - You tag said content in some way which allows you to understand how difficult it is. - When in "review" mode, you essentially choose what to reread based on the difficulty you are feeling has the most impact right now.
One of the frustrating things I experienced trying to do self-study using Anki and self-created "extensive reading" was that I found I was spending a huge amount of my time doing things to enable me to study, rather than actually studying.
So my ideal vision for what the interaction looks like is this:
1. You ask for something to read / watch.
2. It gives you something
3. You read / watch it, looking up words you don't know, going on whatever rabbit trails you need to to understand what it's given you.
4. Repeat.
IOW, I want the algorithm to do the "look for stuff to study" and "decide what to review", so I can just focus on the "study" part.
It would have a huge library of content ("corpus" in the jargon) -- news articles, Wikipedia, stories, social media posts, movies, TV episodes, whatever. The algorithm would select something for you that's at the right difficulty level -- hard enough that you're learning, but not so hard that you get frustrated. The algorithm would arrange for words that you need review on to just appear in the content you happened to be reading.
Maybe you give it parameters (something long / short / fiction / news), maybe not. Maybe you give it feedback about which words were harder or easier than it expected, maybe not.
Certainly a useful feature would be to say, "I want to be able to read this", and give it a link to a webpage or YT video, or a PDF; it could then focus your learning / review towards being able to read that particular content.
But the common case would be that you just come to it and be given a feed, as you would with Facebook and Youtube -- except the algorithm would be trying to teach you a language, rather than trying to get you to watch ads.
This is an intriguing idea; kind of like the method used in "Lingua Latina per se ilustrata" but automatic and with rich media. This is definitely possible, though the domain space is quite large, some tradeoffs would have to be made that might stifle the exact vision.