So I've been doing some brainstorming. I'm thinking that to get the kind of audio ecology I want with an appropriate 'signal-to-noise ratio', it would be something like so: an iPod playlist consisting of Japanese audio segments from your favourite media as Nukemarine mentioned, but divide the media segments into percentages according to their maturity levels. (According to Anki, a mature card is a card with an interval over 21 days.) I think a good way to do it would be ~10% of the audio would be young and/or new, 60% would be youngish mature cards (intervals of 22 to 120 days), and 30% would be very mature (over four month intervals).

You could make these playlists by taking your Anki deck, sorting it by intervals in the card browser, selecting and tagging groups according to the aforementioned percentages and then exporting them--from there either using the cards' audio (if they have them) or making them into longer segments.

Or (and I think this is best) if they don't have audio or if you prefer other versions, you could use those exported Anki cards as whitelists for subs2srs, creating decks from those where the subtitle lines that don't include those exported cards are eliminated, and then extracting and manipulating the audio from there (perhaps you'd want the segments to be lengths that would fit between timeboxing sessions?).

So in addition to 'priming' yourself the standard way by studying a subs2srs deck and then watching the show it was made from, and then cutting up the audio for shuffled listening as per Nukemarine's suggestion (edit: or rather, inserting an audio-splicing step *before* watching was his suggestion), you can make sure your entire audio listening environment consists of a controlled ratio of familiar and unfamiliar cards, the 'noise' accentuating the 'signal'.

I say 'accentuate' because I think that the 'slightly mature' cards are in a transient place in your mind, on the edge of fuller memorization, perhaps limited in their context as they haven't been encountered that often in and outside of Anki, where instead of being overlearned, as long as you shake things up with randomness, a new context, and just a bit of unfamiliar material that knocks you off balance a bit, they're reinforced as per Supermemo's thoughts on planned redundancy. Similarly, the unknown material, being limited as it is, will fall into the i+X (or +1 as you prefer), and the 'very mature' cards are kind of like stabilizers.

What do you think? Sorry if it's a bit mixed up right now, there's so many paths and tools to arrive at the ecology there's probably a few missteps. But you get the basic idea. You could perhaps do something similar by sorting by 'ease/difficulty' or something else rather than 'intervals'.

Still working out the possibilities for text (and images and manga).

I think...


....Don't you agree?

I think this idea has a good signal-to-noise ratio.

How about if the playlist was made from items that were recently reviewed in Anki. This would naturally give you a distribution of new, young and old material of the same ratio as Anki shows reviews each day. For a few days after reviewing something in Anki, you would also listen to it in your playlist to further enforce the review, followed by a spacing period until the card is again due.

vosmiura Wrote:I think this idea has a good signal-to-noise ratio.

How about if the playlist was made from items that were recently reviewed in Anki. This would naturally give you a distribution of new, young and old material of the same ratio as Anki shows reviews each day. For a few days after reviewing something in Anki, you would also listen to it in your playlist to further enforce the review, followed by a spacing period until the card is again due.

Well, and that's I think what smart.fm tries to do with its podcast options? But I think that a typical slice of recently reviewed cards, depending on how you select them, would have a lot more young/youngish cards than the others--or rather, it would be a bit more dynamic, more dependent on how quickly you're adding new cards, wouldn't it? I selected different groups of cards after sorting them by due date, most of them had intervals of like 30 days or less. ;p Plus since they were recently reviewed, while changing the context (or transducing into new forms by pairing them up with new sentences by using subs2srs) could shake things up, they'd still be fresh enough in the mind that I'd be worried about their impact on the spacing effect in such large amounts.

Which makes me think, even the 10% new cards I was thinking about, I'd probably want them to be either not due for over a week, or added and not yet studied--with the latter, you would use the new cards 'to be learned' as part of the whitelist, which would allow for the audio environment to be incorporated into 'priming' oneself before watching a specific show, if one wanted, in the way Nukemarine mentioned (but taking his Step 6 and putting it before Step 4, and thus you could have a maturity-based ecology + episode-specific priming together). I don't know, still working on that part, think I'll keep those elements separate for now.

Also still trying to figure out how to mix this with themed decks (perhaps using smart.fm to design the lists, export them and use them as whitelists), longform text (shuffled tumblr posts w/ images, entries to be skimmed?), and Audio Lesson Studio for imaginary conversations (wish more subtitle files had 'actors' designated, would make organizing those one-sided dialogues easier).

I guess I need to be more rigorous in thinking about the workflow as well. One hour environments, to be discarded after a week in favour of a new cross-section of 60/30/10 cards? If so, might I want to save the old list in order to ensure there are no duplicates in the successive ecologies?

BTW, if anyone knows how I might streamline this process of exporting cards according to their intervals to a whitelist and using that whitelist for creating/editing subs2srs decks, feel free. If you're a bored programmer genius and want to write a magic program that does this, even better.


Hmm, actually I'm thinking I need to rethink this. Too many loose ends, like really common words. Damn it.

Come to think of it, I don't even know how to create a proper include list. I should just delete the whole damn thread. I'll leave it up as a testament to my stupidity.

I give up. Can't think of an easy way to do what I want. Sigh. It's such a compelling idea (for me), too, I was really looking forward to it. *cries*

I mean, I can still do 'the poor man's' version of just selecting the mature cards and exporting them, then listening to the audio outside the SRS. But it's not the same. Need to have an automated method of taking an Anki deck, selecting spans of cards based on intervals, selecting only the kanji words, removing duplicates, separating by semicolon, then using as an include list in order to create new audio with those words.

Final thoughts: Okay, for now I'll just do a poor person's version: Create a deck and use it to prime myself before watching a show, that's the short term. Once the deck's reached a certain level of maturity, then splice it into segments and shuffle in iPod, alongside some unknown Japanese audio. Or just try out Nukemarine's method, since as vosmiura mentioned, recently reviewed cards will naturally have a mixture of intervals (since sentences will have plenty of common words).