Author Archives: adam

About adam

Autodidactic Polymath

Adding a bunch of sentences without getting bogged down

I’ve started studying a bunch of new Japanese sentences lately, but I don’t want to get bogged down because I’m still adding 5 new vocab cards per day and I want to make sure to keep up with my reviews.

Since most of the new sentences I’m adding won’t give me too much trouble, I used the reschedule function in anki to randomly assign intervals between 15 and 45 days.  That way, I only get a few per day and I won’t see them as often as “fresh” cards.  In other words, if I added 150 cards with a 15 day interval, I would only see 10 extra cards per day – which I probably wouldn’t even notice.  Unfortunately, the way the function works in anki, if I set them all to 15 day intervals, they will all start on the same day, which is why I used a range of 15-45.  Now I get a few every day.

Even though I’m using the n+1 optimized deck, for some reason I’m getting a lot of words that I don’t know. Probably because I started out learning in non-optimized order.  Whatever the reason, now I’m suspending those cards because I’d rather study the ones that I can read and understand because there’s a lot of them to study.  I’ll sort out the suspended cards once I have a need for more sentences.  Starting out with the optimized deck would’ve made everything a lot easier.

I’ve also added another filtered deck..  “New Core Sentences”  filters all of the new sentences that I haven’t seen.  Here’s the code: card:sentences prop:revs>0  and I set the deck to reschedule.  This works out well when I have some extra time.  Instead of going over and over my 2 day deck for dubious gains – I go through my new sentences.  It’s basically a filtering operation.  If there are words on the card that I don’t know – I suspend it.  If I know all of the words, but can’t produce the closed word – I fail it.  If I know all of the words, but I don’t understand the grammar – I fail it.  If I know what the sentence means, I pass it.  Since I set my “new interval” to 15% even the failed cards won’t go to a 1 day interval, but somewhere between 2 to 7 days.

The effect of all this is that I get a large number of new japanese sentence cards in play quickly without adding a huge burden to my daily workload.  Also, the cards start their lives with useful study intervals so I’m not spending too much time seeing the same cards too often or not often enough.  I can already feel that I’m learning new grammar and I wish I’d done this earlier.

Progress Report: [436d]::[457hr]::[2295vocab]

Yikes! It looks like I only learned 4 new words in the fast 2 months!  However I’ve been adding 5 new words per day…  So what gives?

Leeches

About a week ago I noticed that there were a few words that I had seen a bunch of times and should be leeches.  I investigated and found that there were a bunch of cards well past the 7 lapse leech threshold so I suspended them all.  I wish I had written down how many I suspended, so I’d have an idea how far that set my count back.  I just hope getting the leeches out of the way leaves me with more time to study the easier to remember cards.

More Adjustments

I’ve also limited my 1 and 2 day cram decks to 70 reps, so cards with anything over that will probably turn into a leech.  I should probably drop this filter or adjust my leech limit, but I haven’t, so I’ve ended up suspending a bunch of cards and only have a net add of 4 in the past 28 days.

My Evolving Lesson Plan

I have been feeling for some time that I’ve been learning vocabulary just to learn vocab and that my time would be better spent studying other aspects of Japanese communication.  So I’ve decided to learn about 200 more(basically so it’s an even 2500) and then put vocab on hold.  Instead, I’ll be studying sentences.  I’ve already started adding a few sentences while keeping up with 5 vocab per day.  I’ll probably add a few more sentences each day for the next month or so until my vocal stands at 2500 and then maybe 1 vocab and however many sentences I can keep up with for the long term.  I’ve done sentences in the past, so I know that I can keep up with a lot more sentences per day then vocab, because I already know the vocab in one direction at least.

Studying the sentences helps reinforce what I’ve already learned and helps me remember the vocab in both directions (English to Japanese and vice versa).  It also helps me getting used to the way words are put together in Japanese.  I’ve written before that I feel studying grammar when first starting out is a mistake, and grammar is best learned by getting as much exposure to comprehensible Japanese as possible – at least at the beginning which is where I still am.

Anyway, in less than 2 months,  I expect to be doing maybe 10 sentences and 1 vocab per day until I’ve caught my sentences up with my vocab.  By that time, I expect that I will be able to follow along with  Japanese TV and children’s books with some difficulty but finally being able to put what I’ve learnt to some limited use.  Hopefully things work out as I expect they will.

Another few Anki Tweeks

I’ve made a few new filtered decks in Anki.  Most of them I don’t even study, but they help me keep track of my progress:

The first filtered deck I do actually study from.  This deck vacuums all of the cards from various decks that are due today into one deck. The code is simply is:due  This is convenient, but it’s not the main reason I made this deck.  At the beginning of the day, I rebuild this deck and I can immediately add and study a few new cards.  That way, I can learn new cards first thing in the morning instead of waiting until I’ve studied all my reviews which is sometimes after lunch.  I feel I’m a little more likely to remember new cards the next morning after absorbing them over a full day.

The other decks vacuum up specific types of cards to give me an easy count of them.  I have one each for vocabulary, kanji, and sentences.  They all go something like this ("kanji deck 1" "kanji deck 2") -is:suspended prop:reps>=1  *Note that the iphone app has a bug where >0 produces zero results so you have to use >=1.

My Current Anki Method

My last post was about tweeking my anki settings, but I’m not sure that I’ve outlined exactly how I use Anki to learn Japanese Vocabulary before so here’s a quick rundown:

Goals

As mentioned other places in this blog, my immediate goal is conversation and listening comprehension as soon as possible.  I’m interested in reading, but not nearly as much or as soon.

My current strategy is to learn and remember as much Japanese vocabulary as possible in the shortest amount of time.  To achieve that goal, I try to keep my accuracy around 90%.  If my accuracy falls too far below 90% I am spending too much time relearning the same material over and over.  However, once I get to 90% accuracy, the additional amount of time required for small improvements in accuracy begins to increase exponentially.  So I shoot for ~90%.

Decks

I’m using the core optimized deck and I’m only doing vocabulary.  Japanese kana on the front / english translation on the back.

I’ve decided to put kanji and grammer on hold for now and memorize as much vocabulary as possible as soon as possible.  This is phase 1.  Phase 2 will continue with vocabulary, but add sentences and listening comprehension cards.  I’m at about 1300 vocab in the  core deck and another 1000 from another deck.  There’s a fair amount of overlap, so let’s call it 1800 vocabulary.  When I get to around 1500-2000 in the core deck, I’ll probably start adding sentences again (I’ve already got ~410 sentences).  But of course I’ll keep adding new vocabulary.  When I added sentences before, I could fly through them rather quickly, so I’m hoping I’ll be able to continue with close to 5 new vocab per day for the foreseeable future.

Reviews

First thing every day, I finish all my reviews in my main Anki deck.  Finishing my reviews every day is the single most important thing and I make sure that happens 99% of the time.  I never think of adding new cards when I think I’ll be time constrained the following week.  Some days turn out to be pretty hectic so not finishing happens at times, but it puts me in a really bad mood, because I’m more likely to forget things I’ve learned recently and that’s the majority of the cards due each day.  On the rare occasions where I haven’t finished my reviews, I make sure to catch back up the following day.

After I finish my reviews, I either add a few new cards to the mix or move on to my filtered decks.

Filtered Decks

My main Anki filtered deck captures all of the cards that I answered wrong in the past 2 days(“rated:2:1 prop:reps<70”).  It also excludes all cards that I’ve seen more than 70 times because I don’t want to spend too much time on any single card.  I’m going for quantity for now so if they turn into leeches, I’ll come back to them later.  Steps are set to “1” so it’s either pass or fail.   I go through this deck 2-3 times every day until my accuracy is almost perfect.

I also have an Anki filtered deck which captures all of the new cards (“rated2:1 prop:reps<20”).  This is similar to the previous deck and captures some of the same cards, but it excludes the cards I’ve seen a bunch of times already There’s usually only a few cards in here, so it only takes a few minutes to go through.  Steps are set to “.25 1 10” so I have to answer it correctly 3 times before it’s retired.  However, if I feel I’ve got a decent grasp on a card I answer “easy”(#4) on a it to retire it early.  The first step being only 15 seconds means that even the most stubborn cards get into short term memory in the first pass or two, and the 10 minute step means that I should be able to answer correctly when I see it in the main filtered deck later in the day.  I usually go through this deck 2-3 times in the first half of the day until I can answer most of them correctly in the “2 day” deck.

There is another filtered deck that I don’t use very often.  This deck catches only the cards that I missed in the past day (“rated:1:1”).  I only use this deck if I’ve added too many difficult cards and my accuracy goes down too much(~<80%) in the main filtered deck.  Steps are set to “.25 1 10” so I have to answer it correctly 3 times before it’s retired.  However, I answer “easy” on a card to retire it early if I feel I’ve got a decent grasp on it.

Deck Options

The settings on my decks have been changed slightly as well.  I have my leeches set to suspend after 7 lapses because I don’t want to spend too much time learning a hard to learn word, when the same time could be spent learning 2 easier to remember words instead.  I’ll eventually come back to the leeches after I’ve learned the easier words.

I set my “new interval” to 15%.  This stops Anki from rescheduling cards that I know quite well back to an interval of 1 day.  Sometimes I miss a card that has an interval of several months and I don’t think it’s appropriate to set it back to square one.  Since the progression of intervals isn’t linear, a “new interval” of 15% probably cuts the actual number of reviews for the card by 50% depending on its “ease”.

And finally, I’ve been answering most cards with “hard”(#2) instead of “pass”(#3).  This has me seeing each card more often than if I pressed #3.  I’ve experimented with adjusting the interval modifier, but it didn’t seem to increase the frequency of the cards that I wanted, but also resulted the unwanted behavior of some cards having the same interval regardless of which button was pressed.  So I press #2 unless I either don’t know the answer, or know it so well that I don’t want to see it very often.

I also spend as long as I need to come up with an answer if I think I know the right one.  I feel like the process of trying hard to remember the correct answer makes it a lot more likely that I’ll remember the answer into the future.   So I act a lot like I’m taking a final exam and make sure to come up with the answer if I think I know it and make sure the answer is correct.  Although, If I feel that I really need to study a particular card, I’ll fail it even if I get the answer correct.

Schedule

I try to add my new cards as early in the day as possible.  Research has shown that we are better learners early in the day when we are well rested, and not as good later in the day.  Also, getting my first exposure early in the morning lets me reinforce them better throughout the day.  Unfortunately Anki won’t let me add new cards manually until I’ve done all of my reviews, but that gives me added inspiration to finish my reviews early.

I add new cards in batches of 5.  Most days I only add 5 because I only have about and hour per day and I can only keep up with my reviews if I only add 5 new cards.  But sometimes I add 10 or 15, but only 5 at a time.  If I add more than 5 at a time, I usually end up going over and over them before I can remember more than a few for more then a few seconds.  So I make sure to commit 5 to short term memory until I add 5 more.  After 10, I’ll probably go through my “new” deck or my “2 day” deck  depending on how many cards in those decks.  I’m shooting for decent(80~90%) retention while increasing the interval for each card (by mixing them with other cards).

After doing reviews and adding new cards, I go through my “2 day” deck and my “New” deck at progressively longer intervals.  As noted, the “new” deck is a subset of the “2 day” deck, so they overlap.  I’m trying to get 90% accuracy on both decks, and the new cards need to be seen more often, so that’s why there’s 2 decks.  If a few hours have passed and I haven’t seen the new cards, I’ll go through the “new” deck.  Whatever helps keeping my accuracy up.  But as the day progresses, I’ll need to study less to keep my accuracy at acceptable levels.

As I’ve mentioned elsewhere, I only have about an hour per day to study.  So try to maximize that short time, while keeping my reviews to a manageable number.

Tweeking my Anki settings

This past weekend I added an extra filter to my main cram deck.  It used to be “rated:2:1” to which I’ve added “prop:reps<70”.  This makes sure that I’m not spending too much time studying cards which are probably going to end up being leeches anyway.  When I added this filter, my cram deck went from 66 cards to 55 which is a 17% reduction.  When I consider that about half of my time is spent cramming, I estimate I can spend 8.5% more time studying cards more likely to stick in my memory.  Hopefully that translates to me being able to add more cards.

Two weekends ago, I consolidated my sentence deck with my vocabulary deck and updated the combined deck to optimal order.  Unfortunately, since I had an overlap between them I ended up with duplicate cards and that threw off my card count for progress tracking.  I’ll eventually figure it out, but It’s hard to keep track of my progress for now.  Fortunately with the new order, the cards I’ve added in the last week have been easier to remember, so I’ve been adding more than normal.

I’m also considering adding sentences (and maybe audio recognition) to the mix again now that I’ve synchronized vocab and sentences.  I won’t have the problem I had before where sentences contained a lot of unknown vocabulary.  *edit:  I’ve unsuspended 77 sentence cards as a test.  I used Anki’s reschedule function to set their initial interval to 30-60 days.  That means that I will only see 2-4 at a time, so (hopefully) it won’t set my vocab study back at all.  And if I answer it correctly, I won’t see it again for 3 or 4 months.  If this works out as expected, I’ll filter in more sentences this what until I’m adding sentence cards along with their sibling vocabulary cards.

Progress Report: [365d]::[377hr]::[2165vocab]

Today is my 1 year anniversary of starting to learn Japanese (this time) and I’m happy to say I haven’t missed a day yet!  I’ve averaged just over 1 hour per day for 365 days and I’ve learned 2393 bits of information (vocab + kanji) although I know that there’s some overlap.

I’ll be back with some more thoughts about this milestone as soon as I have a little more time.

Progress Report: [337d]::[349hr]::[1975vocab]

It’s been 60 days since my last report and I don’t even have 100 new vocabulary words to show for it.  I blame the holidays.

Fortunately I didn’t miss a day of reviews even though I was on a road trip, driving 8 hours on some days and skiing all day during others.  However there was one or two days where I didn’t fully finish my reviews, putting in only 5-10 minutes, but I didn’t get more than a day behind because I caught back up the next day.

Part of the reason it was easy to keep up with reviews is that I used the “review ahead” anki feature and set it for however many days until I got back from my trip.  So one day when I had some extra time, I studied about ~400 reviews, so for the rest of my trip my workload was only 20-40 reviews per day instead of the normal 100-150.

Overall I’m pretty disappointed by the small amount of progress, but I’m happy that I didn’t miss a day of study.

Progress Report: [276d]::[292hr]::[1878vocab]

It’s been 51 days(54 study hours) since the last report.  I’ve learned 244 new Japanese vocab words which works out to less than 5 per day.  I still haven’t learned any more kanji or added any more sentences.

The low total is most likely explained by my suspension of all my katakana cards because I felt that they were not the best use of my time at present.  I’m sure quite a few katakana cards were included in my previous total but were dropped from my current total.  That probably means that my old average of 7 new vocab per day was a little high since katakana cards were pretty much already learned.

At any rate, I’ve found some Japanese music with vocals that I actually enjoy listening to, and there’s actually a lot of words that I can pick up on.  There are a few songs with very simple, repetitive lyrics where I can probably recognize 50% of the words.  This is very rare and there are still songs where I can’t identify more than a word or two, but finally getting to a point where I can see a hint of actual practical progress is encouraging.

Japanese TV is a tiny bit better, but I still can’t do better than catching a few words here and there if I’m lucky.  I really do feel that listening practice is helping though.

I’ve calculated that I probably won’t be able to follow TV shows until sometime next summer when my vocab is up around ~3000.

Progress Report: [225d]::[238hr]::[1634vocab]::[354kanji]::[421sentences]

I started learning Japanese (again) 225 days ago on February 10, 2013 and I haven’t missed a day yet.  At the very least, even if I didn’t add any new cards,  I think I have gotten through most of my reviews every day.

I was not starting from zero, since I had taken classes and studied a bit on my own.  I estimate that when I started this time, my vocabulary was about 350 words, I could haltingly read kana, and understand very basic sentence structure.

2/3 of a year later, the biggest improvement is in my kana reading.  Since I have been studying vocabulary using kana, I can read hiragana very easily and naturally.  Even in brush scripts and stylized fonts, I can read probably about as fast as I can read roman letters.  This isn’t to say that I can read as fast as english because I don’t know the words, but it’s like sounding out words in Spanish I guess.  Katakana has seen less improvement, since most of the vocab is Japanese origin so it’s written in hiragana.

I can recognize a few words if I watch Japanese TV but it’s not really helpful to understanding what’s going on.  I can probably understand  5-10% of the words being said and that might be generous.  I was hoping that by 1600 vocab words, I would be able to recognize more words in Japanese television and music, and get a general sense about what people are saying but that is not the case.  It’s a little bit discouraging to be honest, but I have no intention of slowing down or stopping.  Note that when I say 1600 vocab, that there are probably a lot of duplicate words in there, as I started out with a different deck, so there is some overlap.  That also means that I’m only around 600 in the core deck since the other deck has >1000 cards.

I’ve been listening to the audio for the first 500 core sentences in the car and I can understand about 30-40% of what’s being said.  This is also a bit of a disappointment, but a little more understandable.  They speak full speed Japanese, so if it’s a long sentence, it’s hard to process any of it.  If I heard the sentence over a few times, I would be able to understand maybe 50%.   The audio that I have isn’t in the same order as my learning deck, so there are some sentence that I know I haven’t learned yet, so it’s understandable that I wouldn’t know those.  But it’s driving home the fact that knowing vocab isn’t the same as understanding full speed Japanese.

I intended to learn kanji alongside the vocabulary and add in the sentences.  When I know all the vocab, I can fly through the sentences.  Basically I’m just training on the grammar.  But it gets frustrating when there is a word that I don’t know.  At first, I was looking up the words that I don’t know but it seemed inefficient, so I stopped.  I also stopped adding kanji, so now I’m 100% focusing on vocabulary.  For now, that seems the quickest route to where I want to be.  Since my focus is on conversation, learning kanji is of secondary importance although I will probably filter a few kanji in a little later.

There are 100 days left in 2013, and my learn rate has been 7 fresh vocab/day.  So, my goal is to get 700 more vocab or 2334 total by the end of the year.  That might be kind of difficult since the holidays always put extra demands on my time, so I’ve been doing 90 minute days lately to get a few extras in the bank before the holiday rush.  I also hope to filter my decks of all duplicates, so my count is more accurate and study is  more efficient.