aphasiac Wrote:

chameleoncoder Wrote:aphasiac: I found the same issue. After taking the advice of some very knowledgeable individuals on this forum I am now working my way through Core2k+Core6k vocab only (sorted by kanji frequency of course).

Hmm, interesting. I tried studying vocab out of context before, and found it tough - but then maybe with the Core 2000 images and more immersion it might work.
How many words are you learning per day? Any tips on studying them??

I've been slowly increasing the number as I work out my optimal study methods so right now I'm doing about 45 per day and optimally I'd like to be at 60.

Here's what I've found to be the optimal learning method for myself (warning: long).

I have all the unlearned cards suspended in anki. I do the following steps several times during the day or sometimes if I have time I'll do two batches of 15 but always in the two batches rather than 30 together. My process sounds kind of long but it actually goes really fast.

1. I un-suspend the next 15 or so cards (ordered by created). I like to do entire blocks of kanji so if selecting 16 or 18 will give me a more logical grouping I do that.

2. I select the new cards and do the "cram" action on them (this takes hella long to process so I'm thinking it must copy the whole media directory, but oh well).

3. In the cram session I change the ordering from random back to order created and start going through the cards.

4. On most cards I'll paste the word into jisho.org. I'll use the kanji details link on the word to make sure I have the readings of each character correct (in the case of a compound where I can't easily tell). Now if the word is vague at all I'll use the sentences link too in order to see some context of the word. Just enough that I can understand the nuances.

5. Repeat the audio a few times on a card while kind of thinking over the sentences I saw the word in if it's one of those vague ones. Doing this only during the initial learning seems to be all I need for those tougher ones as I seem to form a good mental link between the definition and the context right away.

6. Hit "1" for the card. I do this for every card since cram mode will reschedule the card to come back very soon. It will usually come back after 2 or 3 cards depending on how long I took. I really like doing it this way as it's these little micro reviews that really cause it to stick. In fact on a lot of the cards I have no idea what it is until it's come back maybe 3 or more times during the session. Always hit "1" though on every card though.

7. By the time I reach the end of the 15 or so cards I've usually had quite a few reviews of about the first 10 so those are sticking but I need a little more on the rest. While still in cram mode I open the card viewer and select all the cards and do a reschedule on them as new cards (this is in cram mode so no effect on actual deck). Then on the study options page I switch it to show new cards in random order. This eliminates any I was getting right just because I was used to the pattern. Now I start going through the cards again, fairly fast this time since I should know "most" of them pretty well, still always hitting "1".

8. Once I've gone through cards at least once in it's new order then I start hitting any other key than "1" on all the cards that I've got down solid. One by one the deck will dwindle to none and I can close the cram deck.

9. Now back in my main deck I have the study options set to "show new cards in random order" and "show new cards before reviews" (in case some reviews come due during my session). I now go through those same 15 cards on my actual deck hitting "2" for each one no matter what (unless I fail it and then I'd hit 1 followed by 2 once I got it right the next pass). This makes sure the card will show up in tomorrow's review for sure (this seems important for me).

10. The next day I always do reviews first of course but since I hit that "2" on those cards yesterday the number 2 option today will cause them to be scheduled at about 2 days which is perfect for those cards that I have to sit and think on. I only use 3 if I get the card right away (like many of the super common or kana ones).

aphasiac Wrote:

chameleoncoder Wrote:To soften things even further I took the 800 or so words in the list that use kana only and spaced them after every 5 words that contain kanji.

That sounds really neat! couldn't have a copy of your deck could I?

The deck is huge with all those audio clips, nearly 320MB. Do you know of any good places to upload that much? Otherwise, you could use the smart.fm importer on the full core2k and core6k to get the audio and I could just upload a fact export.

hey, thanks for the great report!! Nice methodology! you're basically doing an electronic version of the Iverson method, will have to give it a go!

This hopefully will get around the main issue I have, i.e. I cram a word until I remember it, but I know for a fact I won't remember it tomorrow when I hit "2" in Anki..

As for the deck, I've already got the Core 6000 audio - just after your facts as a text file! Use megaupload, I'd really really appreciated it (it's mostly the kanji-kana thing that sounds sweet)

aphasiac Wrote:hey, thanks for the great report!! Nice methodology! you're basically doing an electronic version of the Iverson method, will have to give it a go!

Yeah, I'm not a fan of the Pimsleur material but the method does indeed seem to work for me.

aphasiac Wrote:As for the deck, I've already got the Core 6000 audio - just after your facts as a text file! Use megaupload, I'd really really appreciated it (it's mostly the kanji-kana thing that sounds sweet)

In that case here you go.

http://www.megaupload.com/?d=DUJTGE71

Hi I just found this thread and am interested in getting a reliable KO2001 deck and using this sort method. Can anybody supply the word list as well as how to do it? I've looked through a couple pages, and I can't find the exact concrete method.

I just need:
1. KO2001 sentence file.
2. Sorting method and whatever is associated with this method.

I think I'll work harder on my sentences and kanji if I can get a sorted list. When I first tried going through KO2001 (with my book and manually typing it out as I went) it was excruciating and demoralizing every time they dropped countless new words onto me. Thanks.

chameleoncoder Wrote:aphasiac: I am now working my way through Core2k+Core6k vocab only (sorted by kanji frequency of course).

you could always take a deck that has vocab and sentence together in the fact and sort on vocab only, but still have the sentence there for context. so you can still read the target vocab in context, just don't worry about learning the readings for the rest of the sentence

For those struggling to run kanji-sort (and I did struggle a bit), take a very close look at cangy's post here:

http://forum.koohii.com/showthread.php?p...2#pid77392

A few things to keep in mind: it will work on pretty much any tab-delimited txt file, so long as it's UTF-8 encoded. (Unicode will make it barf errors, as will anything else that ISN'T UTF-8.)

I found using a program like TotalEdit really helped, because I could change the CRs from Windows to Unix if I felt like it under Save As..., and I could also change the encoding to UTF-8. (Which helped.) This way, you can grab spreadsheets from wherever, and convert them to UTF-8 without having to import them in Anki.

Setting my computer to run in Japanese mode also helped me when it was barfing. Not sure if it actually helps it run. (Does it?)

Make sure all of your files are in the same directory as your perl script, of course. And make sure you install perl on your machine if you're using windows.

Most importantly, getting the syntax right is key. This is the bit from Cangy's post that clicked for me.

kanji-sort --kanji kanji-ko2001.txt --sentence-field 2 < exported.txt > sorted.txt

How do you translate this into what to do at the C:\ prompt? From the proper directory, I typed this:

Code:

perl kanji-sort-1.5.pl --kanji ko2001.txt --sentence-field 1 <tanuki_ultra_simplified.txt> output.txt

NOTE: I renamed the file kanji-ko2001.txt to ko2001.txt, to avoid confusion. I didn't have to type any spaces around the "<>" characters

And yeah, as you can see, I just ran it on a very stripped down version of the Tanuki corpus, just for the heck of it. There's a link to the Tanuki spreadsheet in the Google spreadsheets sticky at the top of the learning resources forum. I downloaded it into Excel, saved it as a Unicode-tab-delimited-text file, then opened it in Total edit, and converted it to UTF-8. Takes all of 5 minutes once you figure out the process. (Figuring out the process took an hour or two.)

Okay, my last post was a mess, so I deleted it and rewrote it in English... a giant wall of it. Sorry.

As Nuke and others discussed earlier, tacking on the RTK list to the KO2001 sorting list removes a lot of the oddballs from the core lists... but it doesn't catch all of them.

What I wound up doing after some trial and error was to take the RTK1 and RTK3 lists, sort them with the KO2001 list, and use that for sorting. This picks up all 3007 RTK kanji, and dumps the RTK3 kanji in the back of the line.

Quite a few of the RTK1 kanji aren't picked up in KO, and a a few of the RTK3 kanji are picked up in KO. This also removes any duplicates you would get if you had just added one file to another. (Which is easy, but inelegant.)

I also found the missing KO kanji #1777 (璧) and stuck it back in the file. I left out #2001, which is the repeat mark, because using it to sort makes a mess of things.

I sorted the entire core 6000 deck (2000 + 4000) exported from Anki with that list, and it seemed to sort pretty well... except for a few oddballs, which I wound up adding to the end of the list.

After I sorted it again, I had managed to stuff 210 oddballs to the back of the line. Not bad.

I don't assume to have caught all of them, but I'm pretty sure I got most of them. If you find more, by all means, let me know. We can either add them to the sorting list and re-sort, or just edit the big spreadsheet. Or do both.

Here's a Google doc with the new sorting list for KanjiSort 1.5

This is the ko-rtk1-rtk3-misc.txt file:
https://docs.google.com/Doc?docid=0ATXlC...dGc5&hl=en

I've set it so it can be edited-- if you find new oddball kanji that aren't already in the list, please add them to it. Do a search in the file before you go slipping one in, and slip new ones at the back end of the file.

You should be able to just download it as a text file and use it with kanjisort-1.5.pl. If you have problems, open it in a proper text editor and save it with UTF-8 encoding, and UNIX LF returns.

I've sorted all 5998 of the total core 6000 with the new list, and saved it as a Google doc here. (Very tedious.) You can export it and use it as an Anki deck, too.

This is the total core 6000, sorted with the modified ko-rtk1-rtk3-misc.txt file:
https://spreadsheets.google.com/ccc?key=...PTFE&hl=en

I'm leaving in the two numeric fields that kanjisort 1.5 adds. Also, while the spreadsheet has a link for the audio, it doesn't have a link for the photo. Sorry, the deck I downloaded didn't have photos. If someone has a better idea, run with it.

The downside is that after 2000, the index numbers in the resulting spreadsheet won't help you look up the kanji at all. They'll just be a rough guide.

2001-2087 are rtk1 kanji not in KO, and 2088-3007 are rtk3 kanji not in KO (most of them.) The rtk numbers for rtk3 will be off, so you'll need to look them up individually. Anything higher than 3007 is a random oddball kanji.

In order to help a little, I created a spreadsheet that lists all of the kanji-- the first 2000 are in KO order, then the index numbers change to RTK numbers.

Here's the link: (This is just a lookup table for the textfile)

https://spreadsheets.google.com/ccc?key=...VTEE&hl=en

I realize this probably covers a lot of ground others have already covered, but I had an afternoon free, and I couldn't find a sorted deck I was happy with.

EDIT: I just ran through the Tanuki Simplified with this, and it went through it like butter. Not really sure how to get it up to Google docs, though, because it's monstrously huge. Any ideas? Tanuki is really handy for getting you past 1110, that's for sure.

Rich,

If it'll help, here's the spreadsheet had posted on google documents thread. I added a 2001KO column, so sorting by 2001KO, then by RTK will give you an organized list for the sentence sorting program.
https://spreadsheets.google.com/ccc?key=...relE&hl=en

In addition, if you're using the Anki file I posted, I sort of kanjified a few words prior to posting. Udon and other words will be in kana if they're grabbed from Core 2000. Sorry about that.

Nice thought, but I already used all of those kanji in the new sort table list thingy. The new sort table has all of the RTK1 and RTK3 kanji (all 3007), with the first 2000 in KO order (if that makes sense.) All it needs now is more kanji after the 4 miscellaneous ones I added to fill it out.

I spent all day sorting, and I think I've become a little unhinged in the process. I didn't particularly want to mess with trying to get all of the core 6k off of smart.fm all over again, TBH. I don't mind a few unique kanji. I've seen 薔薇 a lot in books, although I haven't seen 饂飩 much. I usually see it in kana.

Right now I'm just trying to figure out exactly what to do with all of these sorted... things. I finished KO book 1 a while ago, but "life" got in the way, and kept me from getting on with book 2. Now I'm trying to figure out what to do about book 2, so I'm looking at these 3 options:

1. Tanuki simplified. It sorted pretty well. It will take me all the way to 2,000 kanji with some ease, but that will take 5,000 sentences. 2300 just to get from 555 to 1110. Yikes. The lack of English is smart, though.

2. KO Book 2 sorted nicely-- 1500 sentences, much faster than core 6k or Tanuki by 800 or so to get to 1110-- but they're all looong sentences.

3. Core 6k has nice, short sentences but there sure are a lot of them. 2200 to get to 1110 from 555, and not much after that to get me to 2,000-- in other words, I'd have to go to Tanuki or some other source for help. (KiC?)

I'm thinking I'm just going to look at the book, look at the vocab they want to add, then look at the datasheets, and see what sentences I have available, and not sweat it too much.

Rich,

Again, I'm going off what others have posted, but there's merit at a certain point to just go straight vocabulary lists. With Tanuki, that's a good candidate as you can have (Ques - Kanji Word; Ans - Definition + Sentence). Jarvik is just taking things he's copied via Rikaichan then later importing that straight into Anki (Kanji Vocab, Kana Pron., Engl. Def) and says he can get near 200 words a day.

Also, about Tanuki, I did a spreadsheet scrub of it like this: I combined Core 2k+6k and Tanuki, sorted by Kana Vocab with preference to Core series being top order. Next I had the formula compare the Kanji Vocab to the one above, and mark if it matches (looking for duplicates). That took out about 3000 or more entries from Tanuki leaving a few dupes that are easy to ignore later.

Thanks for the info. Something to think about. I haven't had a lot of luck going with just bare vocab. I need a little context. Maybe when I get better, I can go with just plain vocab. I am going with word lists to learn new vocab, but for SRS reviews, I use the vocab in sentences (which are getting shorter every day.)

I'll probably wind up picking and choosing like I always do. Tanuki has a lot of great stuff in it. So does Core 2+6k. Then again, yahoo.co.jp also has some nice short stuff in its dictionary, too. At least sorting those two was useful.

rich_f Wrote:As Nuke and others discussed earlier, tacking on the RTK list to the KO2001 sorting list removes a lot of the oddballs from the core lists... but it doesn't catch all of them.

if you want to sort by a kanji list and have all remaining kanji sort to the end, see this earlier post

rich_f Wrote:I also found the missing KO kanji #1777 (璧)

thanks for that -- I was wondering why my ko2001 list only had 1999 kanji...

Just a quick note: I've updated the sort file on Google Docs here:
https://docs.google.com/Doc?docid=0ATXlC...dGc5&hl=en

It now contains 3383 unique kanji. All of the RTK, plus the new Joyo, the Joyo candidates that failed to make it in, the old Joyo that got kicked out, and some official name kanji. (The 376 extra are the ones that aren't in RTK 1 or RTK3.) Many thanks to Katsuo for pointing the way to good resources on the Joyo upgrade history.

I'm sure there are other kanji still out there that I'm overlooking. I'll get to them eventually. I've looked at the thread on the boards here, and I don't want the file to go into ridiculous territory. (Yet.)

But this sort seems to have picked up everything in your core 2k deck, Nuke. XD

I fixed the Core 6000 spreadsheet on Google Docs with the fixes as well. I wound up finding 3 more sentences and moved them to the back of the line.

Here's the spreadsheet with the fixed Core 2k+6K:
https://spreadsheets.google.com/ccc?key=...PTFE&hl=en

Now that's a useful spreadsheet. It still has the Sentence numbers so it's easy to sort by Core series then by KO series.

What I'm going to do is export a list from Anki with my suspended (ie not studied) cards marked as "resort". Use that with this spreadsheet then sort just the suspended cards to put back into Anki. I had stopped learning vocabulary directly due to the annoyance of Core 6k's order, but this'll be the thing that'll get me back into it. Thanks Rich (and Cangy for the sorting program).

No problem. I got a little obsessed.

I probably could have just used the Kore spreadsheet, but for some reason or another, there was a bunch of HTML in there that I had problems getting rid of, so I just decided to redo it myself.

Oddly enough, when I looked at it again tonight after all of my sorting, the results are roughly the same. (At least they are at the back end of the list.) I'm not exactly sure how the sorting was done for that spreadsheet, but there you go. Weird.

Oh, and probably the quickest and dirtiest method would be just to use one of the spreadsheets that has a dump of Jim Breen's KanjiDic (I've seen it on GoogleDocs somewhere) and sort it out against the current sort list and then slap the remainders onto the list... but then you're talking about something monstrously huge and frightening... albeit thorough.

Either way, now you can take apart any deck you come across and sort it KO style.

And at least this has inspired me to finally get serious about picking up some PERL knowledge.

cangy Wrote:if you want to sort by a kanji list and have all remaining kanji sort to the end, see this earlier post

that is, generate a frequency list for the file to be sorted, and append it to the end of the sort list. no need to make a list of every possible kanji. here's an example of doing so

kanji-sort is a generic unicode character sorter and doesn't know anything about kanji (despite the name!), hence the current behaviour. if generating the frequency list and appending it is a problem though I can add an option to sort all additional kanji to the end

Now that I know how regex works, I could have saved a lot of work by exporting Kore as a Unicode file from Excel and regex-ing the hell out of it in EditPadPro. Probably would have taken 2-3 minutes, tops. >_<

Learning regular expressions has been a huge eye-opener. I was putting together some audio lessons to listen to in the car on the way to a review session with Audio Lesson Studio, and I needed to mass-change some file names in a .txt file... piece of cake.

I get what you said about generating a freq list now, but I'm not going to be sorting much anymore anyway. I've got 3 databases full of stuff to deal with now-- 4 if you count 完全マスター２級. Not that I'll add all of that, either. I'll pick and choose. But I have a common frame of reference to work with, and that's what I wanted more than anything else. (That, and no weird kanji in my sentences.)

Lol, after reading this thread (and the core6000 thread), I knew for sure that I'm doing something completely different than most here: I'm learning Core 6000 in randomised order (With the randomise function in Anki) It stood that way when I downloaded the deck and I thought, yeah what the heck, why not try it randomised.
My idea was (and is) that if I learn the cards randomised, that I will learn all the readings better, because if you learn many words in a row with the same Kanji, you don't get a 'smoothed out' exposure to the Kanji. I have to note though that I had some vocabulary to begin with: this is probably a not so good idea for a beginner.

Ow and I'm doing KO2001 and Core6000 at the same time I like vocab decks way more because it is so much faster and efficient to learn new words, but KO has the nice sound sentences. I agree though that KO2001 is a bit too bussinnessy.
To complete, I also have my own vocab deck, composed of words I encounter with strange kanji and colloquial expressions.

Call me crazy

Kame, it's not crazy. The original order of Core 6k is pretty much randomized till you get near the end when it's in Kana order. A few have done it in that order. A true random order like you're doing is not much different.

However, many have posted about the ease of learning when using an order like the one from 2001KO. This happened again when we made the 2001KO list on Smart.fm. Later, Cangy showed his list which organized the 2001KO sentences themselves by kanji. Then came the next logical step of organizing Core 2k/6k. Then it's a matter of sub-grouping those (beginner, basic, intermediate).

Basically, this "Sorted Core 2k/6k" is a new thing that most have not done. We're just getting good reaction to it, so it's quickly becoming a standard suggestion for those at that level.

For random exposure, I'm going to leave that to subs2srs. These systematic lists I prefer to be a bit more intuitive when it comes to learning.

After a month and a half of SortedCore2k, with 1524 cards done, I'd strongly recommend it to anyone studying general vocabulary. Seeing a new kanji several times within a few cards, usually with the same reading in each, has helped me get over my difficulty picking up on-readings. My reading ability has exploded. Right now Anki estimates about 99% retention, which, considering that about 50% of the words I've gone through were new to me, is fantastic.

I'm using a sort of modified Iverson-type method to get words into my head. I go through 10-40 new cards a day and write new words in random order on a piece of lined paper folded in half. If a word is interesting, I look it up in a J-J dictionary and may add that definition to the card. I write down on the paper the word in kanji (two lines high), and the reading and meaning under it. (For hiragana words, I write it in English with the kana underneath, since there's usually one kana word a day and I'll immediately remember that word without even reading the kana...)
After I'm done with Anki, I read down the list, covering the reading/definition, until I can do it smoothly. Then I do it another two or three times in the evening, during commercial breaks and the like. It's only a few minutes extra work and it improved my retention to the next day immensely.

Been reading through this topic and i am still not quite sure how to reorganize my KO2001 deck based on the graphs that cangy showed. Been using Anki for my RTK deck but its pretty simple and iv never done any of this complicated stuff with it. Could someone write out and Idiots guide to reorganizing my anki deck to make it resemble cangys 2nd graph or redirect me to someone whos already posted it or something?

*push*
I really would like to cover the rest of the remaining Joyo-kanji (again: frequency orientated. Learning kanjis that appear comparably lesss often) after core6k, I will be finishing in exactly 47 days (hurray!). Is there a deck/option to sort the words out of KO2001 I already learnt in core6k? If so, please tell me how/give me a link please! Thank you!! :-)