[For other folks reading along, scivola is experimenting with using native materials very early on, and SRSing both his native materials and his courses. This is somewhat experimental stuff, and only a handful of people on the forum have tried it. So in the post that follows, I'm going to dive into some nitty gritty details, both to compare notes, and to alert him to various obstacles that I had to work around.]
scivola wrote:So far, I'm enjoying this method very much. Today was lesson 44, so I'm far enough in that I've got "passive wave" cards in all stages of disappearing, and a decent number of "active wave" cards showing up. I didn't mention that for the active wave cards I've got it set up so that I have to type the L2 sentence, and I think that is going to be a big plus. It has forced me to focus on all the various accent marks that I wasn't really aware I was glossing over until now.
Ah, interesting!
Personally, I've found that if I make my Anki cards too hard, my failure rate goes way up, and I run into a huge number of "leech" cards. "Leech" cards are the ones that I fail again and again, causing them to be recirculated endlessly through the first few weeks of reviews. I find that these cards tend to be both miserable and highly inefficient. (The highest efficiency output cards for me are probably MCDs, which require only a single piece of output, and which are stupidly easy.) If I find that my reviews are getting really difficult, with lots of leeches, then I take two steps:
- Whenever a card appears and I groan, "Oh, no, not that card again," I force myself to immediately delete the card. It's OK to have a tiny handful of cards like this, but if I have a lot, then I just delete the groan-worthy cards individually during reviews.
- If I'm deleting a lot of cards, I try to make my cards easier. For example, if I were struggling with a bunch of "See L1, type L2" cards like you're using for the passive wave, then I'd start gradually replacing them with cloze cards where I saw all the L1 & most of the L2, and I only had to type in a single "interesting" word or syllable.
So that's my process of dynamic adjustment: Delete the unpleasant cards and the leeches, and replace them with something easier. As Khatzumoto
wrote:
Khatzumoto wrote:Waste products — bad cards — are a natural results of the “metabolic” process of SRSing.
Khatzumoto ran a very expensive "course" that was based on native materials and SRS, with a money-back guarantee. (
UPDATE: It's possible that AJATT has largely shut down as a business.) So he saw a lot of people using techniques like this, and he collected a lot of data about where things went wrong for particular groups of students. And as he iterated on the course design, he wound up moving to easier and easier card formats, and he started strongly emphasizing card deletion. I followed a similar progression with my various French/Egyptian/Spanish decks.
For a first-hand view of what a later version of Khatzumoto's course looked like, see
Rapp's Neutrino log on the old site (short and excellent reading). For examples of how Khatzumoto simplified his recommended card formats, see the
lazy Kanji post and the
MCD posts. As always, feel free to ignore Khatzumoto's hype and just dig for the data. And again, I'm not necessarily recommending that you adopt any particular card format—there's still room for lots of experimentation there. Rather, I'm pointing you towards "case studies" of other people who have experimented in this space, just in case any of their insights prove relevant at some point in the future.
So if what you're doing seems to be working, don't worry about it, and please write about your experiences for the next generation of people who try this stuff.
But if you ever feel like you want to adjust some details, then some of the posts above might provide helpful insights.
scivola wrote:That does bring up a question I had for you and/or sprachprofi. In her article about learning to understand your favorite tv series in 30 days, she mentions that she created filtered decks from the original overall deck that subs2srs created, picking out short, simple sentences to learn first. I don't know if you followed that method or not. But if so, I wondered if you had any advice for balancing picking simple sentences from additional episodes versus going back through the episodes I've already created decks for and picking out slightly more complicated sentences.
So, two things to keep in mind: (1) At this point in her life, Sprachprofi is a hard-core language-learning badass, and I can't reproduce the results she got with Japanese. I can get close enough that I
believe her results, but I'm maybe about 25% as efficient. Which is still pretty mind-blowing, actually. (2) Sprachprofi was working on a totally unfamiliar language, with no "related language discount." So for her, searching through the deck for short, closely-related sentences was probably the only way to jump start the process.
In my case, I'm a native English speaker with C1+ comprehension of written French, which adds up to a pretty substantial "discount" when learning Spanish. I initially started with the movie
Y Tu Mamá También. This was fairly challenging, but not as challenging as tackling Japanese (or ancient Egyptian) would be. So instead of searching through the deck, I just started from the first card,
but I deleted ruthlessly. Counting both initial reviews and reviews in the first few weeks, I think I deleted at least half of the cards.
Later, I switched to
Avatar: la leyenda de Aang. This was a lot easier, and I'd already learned the basics from
Y Tu Mamá También. So this time, I just cranked straight through and I only ended up deleting about 10% of the cards.
scivola wrote:I would say that I understood maybe 50% of The Little Prince. I read that about a month ago, right at the start of this attempt at French. A lot of chapters were much better than that, while several of the last few were more like 20%.
When I first attempted
The Little Prince, I was probably about B1, and I used LingQ (today I'd recommend readlang instead). I made it through several chapters, and it was a struggle. Later on, after passing my B2 exam, I tried again, and the book was comfortingly easy to just sit down and read. So I'd guess that your comprehension of written, formal French is maybe somewhere roughly in the B1 range, especially if you're getting "much better" than 50% comprehension on at least some chapters.
scivola wrote:The Aldebaran BDs haven't had that large variation. I have definitely noticed that I would learn a grammar point or some idiom either from Assimil or from my Buffy subs2srs deck and see it show up in the BDs. So that's encouraging. I would say that I have understood maybe more like 65-70% of those. My reading speed is glacial, sometimes around 10 minutes per page, and comics aren't exactly the most text-heavy things. Some sentences I just don't get at all. But most of them, if I re-read and kind of puzzle it out, I can at least get the gist. I'm definitely understanding the plot lines well enough that the story itself is intriguing, which provides a lot of incentive to keep studying/reading.
Very cool! From this description, I get two things:
- When reading in another language, you're unusually tolerant of ambiguity. The linguist Stephen Krashen reports that some students require about 98% comprehension before they're able to enjoy extensive reading. But when I read Toqueville's parts of De la démocratie en Amérique, I had a total blast, even though my comprehension was much lower: I understood maybe 60% of the text fairly well, and I could puzzle out another 30% if I stared at it long enough. The other 10% remained opaque even if I used a dictionary and stared at it for a while. Similarly, when I started watching Buffy extensively, I could follow maybe 40% of the dialog (though I could understand far more if I read a transcript). But for me, once I reached that level, things started to "take off", and I could learn rapidly just by reading and watching.
- You're having a lot of fun and you're enjoying the story. This is my personal litmus test: If I'm having a blast, I never worry about the difficulty level.
The nice thing about Aldebaran is that it uses what I think of as "core" vocabulary, and there are lots of books (19 across all the related series). Basically, nearly all the vocabulary and expressions you see in Aldebaran will reappear elsewhere by the time you've read, say, 5 science fiction or adventure novels and watched a couple of TV series. Plus, Aldebaran will give you a strong base of spoken French.
scivola wrote:On the Buffy episodes, I can definitely understand the sentences that I have studied through subs2srs. I can pick out bits and pieces of the sentences I haven't studied, but it is tough. It mostly seems to be a matter of attuning my ear to the very fast speech and omnipresent liason. I have many times had the experience of not quite being able to understand a sentence while feeling like I should be able to, and when I check the transcript and see how it is written I can understand that readily, but the actual spoken sentence sounds nothing like I would have expected.
Yeah, some of Buffy's audio can be pretty challenging. Maybe 40–70% of the dialog is relatively straightforward once you're familiar with the series, but 5–10% is pretty hard even at B2/C1. So as always, suspending or deleting the annoyingly difficult cards is a good idea.
Overall, it sounds like sentences are sticking nicely once you learn them, but that you're not seeing that comprehension transfer to unfamiliar audio yet. Normally, I would expect to see that transfer start occurring once I had, say, at least 750 cards that were at least 20–30 days into their Anki review cycles. At that point, you should be starting to understand some cards the first time you hear them. If you're not seeing much transfer at that point, you're probably in unknown territory. At which point, I guess, it's time to ask the usual questions: Am I having fun? Do I feel like I'm learning something? If the answers to those questions are ever "No", then
Rapp's advice is excellent:
Rapp wrote:And for the cards themselves, I'm much less attached to them. Surusu makes it very easy to generate scads of cards in Khatz's MCD format. Having to invest so little effort in creating the cards makes it easy to delete them if they are in any way confusing, badly formatted, or whatever. So I delete lots of them. That has lead to a mindset where I know that no single card is critical to my eventual fluency, so if I decide it is not an excellent card, it gets deleted. If I have read one paragraph from a book and feel like I don't want to continue for any reason, I stop and do something else.
The critical thing is to do something in Spanish frequently. But no particular "something" is important enough to suffer through. There's always a different "something" I could do that would be fun. So I just go do that instead. Easy peasy.
I've been emphasizing this idea pretty strongly throughout the post, because it's a valuable troubleshooting skill for people who focus heavily on native materials and Anki very early in the learning process.
Basically, I think about this process as having three phases:
- Bootstrapping. This is what you're doing now. The basic idea is to claw your way to the next level as quickly as possible. Assimil, subs2srs, etc., are all very useful tools at this point. You're well into this process, and you're definitively getting near the next phase. (I estimate that you're reading between 600 and 900 words per hour in French right now, with enough comprehension to have fun.)
- Snowballing. This is amazing when it starts, and you can find a classic example in sfuqua's log over the last several months. Somewhere between, say, your 500th page of reading and your 2,500th, your reading speed will go up dramatically (to maybe 5,000 words per hour), and you'll go from slogging through kids books to reading actual adult books comfortably, with only a handful of unknown words per page. Similar transitions will happen with TV.
- Polishing. This is where I am now, and it's nice. I assume I can read pretty much anything I want, that I can go into a French movie theater and enjoy many movies without subtitles, and that I can even take an online, university level course for native French speakers (if I want to work myself to death). I still have a lot of work to do, of course, but if you told me I was going to start working in a French-speaking job tomorrow, I'd look forward to the challenge.
Anyway, this was a long post, and you should totally feel free to ignore any part which doesn't seem helpful at the moment. I only wrote you such a long post because you're trying out some pretty experimental ideas, and I want to make sure you have the full collection of diagnostic and troubleshooting tools required to get out of any sticky spots.