nest0r wrote:

Tobberoth wrote:

Think about it. Adding extra crap to cards is the same as adding readings, example words and tons of other stuff to your RtK cards. You think you're hitting two birds with one stone, when in fact you're throwing 30 stones and not hitting a single bird.

I've already thought about that. It's about using different modes to reinforce/augment the encoding process, to simplify the complex into robust cues. With RTK you're using muscle memory, narratives, imagined stories, and keywords to get these complex kanji with the appropriate spatial locations correct and review them till it's internalized, making adjustments to the process, tweaking stories, writing the kanji as you review if you need to, as you go along till you've 'mastered' the kanji.

With other types of information you can do the same thing, except since it's the language itself occurring in context, with audio and text and meaning, there's plenty more you can use to enhance the encoding of the memories, and the goal for me is to find the optimal way to combine them all so I can organize them all into these efficient retrieval structures and parlay them into linguistic knowledge. Rather than info overload/interference, it's synthesizing and enhancing memories. Everything I've read and continue to read about how the memory works tells me I'm right. Also, 9 months of doing this and honing it tells me it works. ;p

I don't think it's the same thing at all. But regardless, have you ever thought that you might be cheating yourself? When I see 裁判 in my cards, I hear the word in my head. I see a picture in my mind of what it is. This is automatic, it means I understand the word. You don't. You have it spoken on the card. You have a picture of it represented on the card... so what is your part in all of this? Do you actually learn the word, or are you relying on crutches?

Maybe, instead of creating those connections in your brain, you're letting your card carry the connections for you. Like people who put the kanji story on the kanji question side.

Tobberoth wrote:

nest0r wrote:

Tobberoth wrote:

Think about it. Adding extra crap to cards is the same as adding readings, example words and tons of other stuff to your RtK cards. You think you're hitting two birds with one stone, when in fact you're throwing 30 stones and not hitting a single bird.

I've already thought about that. It's about using different modes to reinforce/augment the encoding process, to simplify the complex into robust cues. With RTK you're using muscle memory, narratives, imagined stories, and keywords to get these complex kanji with the appropriate spatial locations correct and review them till it's internalized, making adjustments to the process, tweaking stories, writing the kanji as you review if you need to, as you go along till you've 'mastered' the kanji.

With other types of information you can do the same thing, except since it's the language itself occurring in context, with audio and text and meaning, there's plenty more you can use to enhance the encoding of the memories, and the goal for me is to find the optimal way to combine them all so I can organize them all into these efficient retrieval structures and parlay them into linguistic knowledge. Rather than info overload/interference, it's synthesizing and enhancing memories. Everything I've read and continue to read about how the memory works tells me I'm right. Also, 9 months of doing this and honing it tells me it works. ;p

I don't think it's the same thing at all. But regardless, have you ever thought that you might be cheating yourself? When I see 裁判 in my cards, I hear the word in my head. I see a picture in my mind of what it is. This is automatic, it means I understand the word. You don't. You have it spoken on the card. You have a picture of it represented on the card... so what is your part in all of this? Do you actually learn the word, or are you relying on crutches?

Maybe, instead of creating those connections in your brain, you're letting your card carry the connections for you. Like people who put the kanji story on the kanji question side.

Yes, gone over this topic before on the forum as well, it was a crucial part of developing my sense of grading and designing of the cards. No, all of that information is not a forced crutch that intrudes upon every review, depriving me of agency. The way the cards are designed and how I attend to them, they enhance the initial learning process and reinforce the target to be memorized as I go along, using or discarding them selectively as I see fit.

In fact, 'my part in all of this' is where my perspectives are stronger than yours, in how I use the SRS as reference point that relies on my self-awareness, continual and dynamic, in optimizing it. You seem to be more interested in passively responding to the bits of information contained in a black box, a rigid vessel, then transferring them to the language as a separate thing, all of this, I would guess, based on your preexisting familiarity with the language accrued in Japan/in class. I think perhaps your 'curse of knowledge' makes it difficult to imagine learning swaths of the language from scratch.

If I'm using an image and audio when learning vocabulary, and my goal for that word is pronunciation/reading/spelling, then I use the image as Question-side cue for the meaning when I need to, if I'm doing recognition I'll make sure I can fluidly subvocalize and pronounce the text and feel solid on the kanji/spelling, then flip it and use the audio to get immediate feedback on those things, and if I'm unconfident in the meaning, then I'll check that as well on the Answer side. The image has already been used, the meaning internalized, and thus having it on the Question side doesn't become a crutch--if I need it it's there, but otherwise it just enhances my certainty (if I even look at it, or perhaps I'll look at it and use it for another word if there's multiple new words--the image isn't a direct one-to-one relation of concepts so it's flexible). Likewise with the audio as a feedback mechanism. This of course constantly varies depending on the goals and types of cards, and the senses always are designed to interact in these reinforcing rather than interfering ways, their presence dependent on my selective attention. Likewise with muscle memory, I don't always write, but if I want to use it I can, so I keep pencil and paper nearby or just use my finger or my mind.

Retrievability and stability. Optimal encoding of the memory when learning it using various modalities for the best retrievability, and using the spacing algorithm and one's self-awareness in grading according to the spaced reviews, dynamically tweaking the presentation of the cards and how you interact with them, to maintain the stability.

PS - To me, this has become super moot now. First there was deconstructing sentences, and finding ways to enhance them. Then transferring video to a similar format with subs2srs; now we don't even have to bother, you can just directly cut up and deconstruct the video with the benefit of the SRS control/spacing. In fact, I think I might retire nest0r. It's the dawning of a new era *cue dramatic music*.

Last edited by nest0r (2009 August 25, 9:32 pm)

cb4960 wrote:

Interesting series of posts! Perhaps subs2srs can be slightly modified to allow for enhanced context and provide a more straightforward approach to making "question/prompt -> response" style cards. I'm thinking maybe something as simple as having an option to add the previous and/or next lines to each fact in the import file.

So subs2srs would be able to output something in this kind of format:
[Audio] [Image] [Text] [Previous line's Audio] [Previous line's Image] [Previous line's text] [Next line's Audio] [Next line's Image] [Next line's text]

This could be good, the next best thing to getting those card-chains I wanted. I'm wondering if there wouldn't then be a way to format a clear visual difference between the previous/current/next sections with colours or somesuch?

IceCream wrote:

That would be Seriously amazing!!! Thankyou!!! Nestor, i don't think you need to seperate it by colour since when it's imported into anki, each one will be imported as a seperate field of one fact (if im understanding right? i'm not too technical). Then you just delete the parts you don't need to make your cards from there. It would save hours from what it takes doing it by hand, and a lot of effort from copy / pasting from normal subs2SRS decks...

Maybe I was imagining something different: I was thinking of having the card pop up, you've got the 'previous' line that you can arrange on the Question side along with the 'call' of the call-and-response, then on the Answer side there's the 'response' and the 'next' line... if that was the format, then automating a formatting process to differentiate those lines could be good. I dunno, I'm hella bad at this technical (or logical) stuff too.

Maybe there's way to change the visuals for specific facts in Anki per deck already... ? *facepalm* Yes, wow, there's this thing called 'fonts and colours' where you can edit the fields.......

Last edited by nest0r (2009 August 26, 12:23 am)

blackmacros wrote:

you're doing a lot of facepalming lately nest0r. Watch you don't take out an eye or something...

Orz

Now you guys are talking about connecting cards in a row in the SRS? That's even more illogical, and the reason to stop putting it in an SRS and just watching the source becomes even more to the point. Having cards in a row is like having all the kanji with the same primitives in a row, too much help, too little remembering, nothing gotten from the process.

ruiner wrote:

Tobberoth wrote:

Now you guys are talking about connecting cards in a row in the SRS? That's even more illogical, and the reason to stop putting it in an SRS and just watching the source becomes even more to the point. Having cards in a row is like having all the kanji with the same primitives in a row, too much help, too little remembering, nothing gotten from the process.

It's just a matter of capturing the information you would find helpful/relevant so you can stick it in the SRS and learn it for retrieval and study it for stability. If the source material is audiovisual, then I also want the material to be studied to be that way, so they're as close as possible. With video, 'close' can be 'exactly the same'. Having the text allows for the bonus of learning how the written language should sound (so you can write with proper nuance and tone), and a map for breaking up the sounds when you study.

For me when I talk about card-chains, if it helps, think of the cards within the chain as words in a sentence or facts on a larger card. You're still (macro-)SRSing it with the aim of being able to instantly parse and comprehend the source materials or materials very similar to the source.

As for kanji, we first learn the kanji in RTK with similar primitives because they help organize our mental hooks, but during actual SRSing they'd interfere with one another if we chained them together, because the target information from card to card would be too similar (well, this is offset by active recall). A dialogue isn't a series of similar primitives, and it exists in multiple sensory dimensions, so you can have reinforcement rather than interference. You can have one sense act as an active recall cue for another, too. Card chains in this sense aren't about clustering similar information so much as preserving the information 'in-between', the same as using a sentence to understand the usage of individual words (and collocation eases memorization and is practical to memorize).

Maybe so, but a sentence can be put on one card, this solution can't, and that makes a huge difference. What if you have a card chain of 5 cards and fail one. Will the whole chain be failed? Will it be broken?

Seems like an abuse of the SRS system to me, if I'm understanding the feature you're describing correctly.

IceCream wrote:

actually, i'm not sure if card chains would really be helpful since anki doesn't support any more than 2 sided cards. If the line is really irrelevent but a few lines later there is a relevent line, you can just change the card manually. hmm, sometimes, if i'm doing something a girl has said and a guy says something important in between, i keep it on the card but ignore it apart from listening, but tbh, it kind of interferes, i've found.

I think what you want though is perfectly achievable in audio studio... i'm not entirely sure how to use it, but can't you just edit the subtitle file, put it in another text file, add in the names of who says it, choose your spacing, then import the resulting sound file as one card into anki? Or simply with audacity, chop up the dialogue how you want and insert blank time...

I think, Like Thermal was saying before, anything that you learn exactly exactly will tend to come out more often, so you kind of have to be selective what it is you're learning to respond. If you learn whole chunks of dialogue off by heart, hmm, it doesn't seem so selective. If you just want longer shadowing, i think either of the methods above can deal with it...

also, i kind of think that it's kind of very different from having double sided cards which help with placing you back at the scene to processing a chunk of dialogue. But, it might just be that i don't need that level of context right now...

Well, that was originally what I wanted for imaginary conversations (http://forum.koohii.com/viewtopic.php?pid=67254#p67254, http://forum.koohii.com/viewtopic.php?pid=67383#p67383)--call-and-response (double-sided, but like Tobbs I think that's a bad name since they're already double-sided ;p), but in lengthier sections of dialogue chained together and studied with the more efficient spacing schedule, extended roleplaying to help with immersing yourself in the context and be able to get used to following along increasingly lengthy pieces of give and take interaction, and these sections as flashcard chains would allow you to mix things up to simulate more off-the-cuff conversations. Something to allow for controlled progression towards 'real' conversations for those who are shy/antisocial/whatever and want to progress that way.

However, since you couldn't do that in Anki, chain cards like that, I was thinking the next best thing would be Audio Lesson Studio/audacity/et cetera. Anyway, I think having the previous/next lines on the cards is a step in the direction I want. But now I'm off on tangents related to video.... + Yes, being selective is key, I agree, but you can be creative in assembling them for spontaneous future use (http://forum.koohii.com/viewtopic.php?pid=68040#p68040).

I'm actually going a bit in circles here, I suppose:

http://forum.koohii.com/viewtopic.php?pid=42730#p42730

Last edited by ruiner (2009 August 26, 11:13 am)

ruiner wrote:

[...] For video dialogue cards, I was originally thinking of just marking down the timings and generating clips based on those spans, [...]

That's what I had imagined too. Simple. Not even for SRS necessarily. The rest strikes me as overly complicated and would clutter up your cards. (I didn't read all of it) And as you say, not all sentences warrant such treatment.

Isolated sentences are good enough for people at the stage of deciphering sentences for vocab, sentence structure and expressions. At some point, however, more complete dialogues can add other layers of meaning and expose the rhythm and tone of conversation. A collection of short video clips that can be reviewed (without having to rewatch the entire show) seemed like a good option. I figured one could use a basic video editor to capture the parts that seem useful. Maybe clips of people to emulate or scenes of, for example:

    * young guys chatting                   * person giving speech
    * couple fighting                           * family members talking
    * strangers talking                        * lovers talking
    * office boss/employee                  * negotiating a mortgage...   
    * seducing unsuspecting women      * apologizing...   hehe

More than just style and expressions, some full dialogues will reveal more subtle uses of language and nonverbal cues that might not be apparent in isolated sentences.  Some of which we might only understand when we view it again after a few months.

I laughed at your comment about learning non-verbal communication from anime. It reminded me of an article I read on how babies develop an ability to interpret facial expressions. The concern was that botoxed mothers would raise babies without this ability.  !?

[edited the unnecessarily complicated clutter of necessaries]

Last edited by Thora (2009 August 26, 10:09 pm)