So I'm curious, how do you folks use 'cloze deletion' in Anki (or outside it)? I am thinking of using it for light novels somehow, perhaps selecting a swath of sentences per volume and SRSing them (with Question side having sentence in cloze deletion form, word on the Answer side), or perhaps SRSing only the extracted vocabulary and then reading the novel normally (but with the SRSed words removed).

You should not check out raseru's thread, because you might accidentally stumble upon ways of unwittingly finding archived batches of light novels via Google.

I haven't used an SRS in a long time, but I used to create a close deletion flashcard in Anki at times when I was reading and had an Aha! moment where I was able to figure out a meaning of an unknown word from the surrounding context. Usually when you're able to infer a word's meaning from context, it also means that you'll be able to produce it when it's presented in this format. In cases where it's still too open-ended, you can provide the first syllable as a hint.

Cloze deletion is tricky because the deleted word has to be such that the context can -only- produce that word in that position. For most speech and writing, this isn't the case. It almost demands sentences made specifically for the SRS.

I haven't done it for Japanese (because I don't see how it's possible, except in very, very specific situations) but I've done it for studying for an exam. It was tough, and I don't really recommend it.

wccrawford Wrote:I haven't done it for Japanese (because I don't see how it's possible, except in very, very specific situations) but I've done it for studying for an exam. It was tough, and I don't really recommend it.

This

Unless have it where you only have a specific amount of words to choose from (like on exams) or you are doing kana -> kanji, then cloze deletion is pretty tricky to work correctly.

For example...
６５歳になって、会社を「たいしょく」した。

６５歳になって、会社を退職した

This way, you can tell that it's 退職 without accidentally writing 大食,　退色, etc...

edit: thanks for catching the させた／した thing...
I know I was wrong, but originally I was thinking of the forced retirement at 65, so it'd be the company making him retire, as opposed to him retiring

But like I said, when the question is ambiguous, give yourself hints to lead you toward the answer without giving it away.

６５歳になって、会社を　＿＿＿＿　した。　(<-- not させた, btw)

may leave too many possiblities, but I don't think this does:

６５歳になって、会社を　た＿＿＿　した。

JimmySeal Wrote:But like I said, when the question is ambiguous, give yourself hints to lead you toward the answer without giving it away.

６５歳になって、会社を　＿＿＿＿　した。　(<-- not させた, btw)

may leave too many possiblities, but I don't think this does:

６５歳になって、会社を　た＿＿＿　した。

If you're going to add stuff to give it away though, what was the point to begin with? Too little help = Impossible to narrow down an answer. Too much help = Too easy and won't train anything. One needs to find the perfect balance but personally, I don't think there is such a thing when it comes to cloze deletion, which is why I don't use it.

I'm using Cloze delete for grammar. For the last few months, it was using it with English descriptive phrases in the closed off part while the rest was Japanese. This got fairly annoying in the more abstract parts later on.

Now, I'm cloze deleting the kana portion of the words being acted up by the grammar point. I even cloze delete a blank space when no conjugation is supposed to happen. So, it's recognition cards, but with a bit of production thrown in to boot. On occasion, I might add a descriptor.

大人[...]らしくするつもりだったのに、大騒ぎしてしまった。
全部飲[...]も大丈夫だよ。
代わりに行[...give to other]。
楽[...]ば、私も行く。

For vocabulary, I'm creating clozed deleted cards for the verbs. Based on the sentence, I just have to figure out the proper conjugation and verb if there's a active or passive type (my main reason for doing this). It's easier in Japanese since kanji is giving you the basic meaning most of the time.

ここに並[...]ください。
赤ん坊が笑[...]います。
靴に泥が付[...]います。
母は味噌汁に水を足[...]。

Tobberoth Wrote:Too much help = Too easy and won't train anything.

If you think providing one syllable of a word is too much help, then I guess you and I disagree on what qualifies as too much help/won't train anything.

Quote:what was the point to begin with?

It's the only approach I can think of for drilling production without incorporating another language. If you've got any other ideas for accomplishing that, I'm all ears.

Thanks for the replies. Seems there's a lot of possibilities. I've tossed a few ideas around over the past cpl years but haven't actually used them .;p The way I've used cloze deletion with other subjects is while I'm reading, I copy/paste a passage into the SRS, cloze delete a few salient points so that the single words are on the Answer side... the goal was only partially to remember particular 'word-concepts', mostly I was just trying to anchor the ideas to one another, because the side-effect of that process was almost a rote memorization of the word when I read the swath of text I took it from.

So then when I was looking at how useful it was to offload goals for sentences (ie the listening/parsing in a video clip deck, the vocabulary in a separate deck), I wanted something similar to practicing reading--except for reading the emphasis wouldn't be on listening and parsing at someone else's pace, it would be to try and get used to reading multi-sentence areas of text without stumbling, getting used to a less strict internal process than I would have with the more intensive recognition cards I did with smart.fm/KO2001 starting out.

In that sense, I wouldn't be doing cloze deletion for reading to memorize vocabulary so much as to get used to the process of associations words with one another quickly, making inferences on the fly, etc., and the 'cheating' aspect of memorizing stuff by rote and having those 'hints' would actually ensure a certain amount of stability, as long as I maintain a smallish pool of words and paragraphs. Anyway, the idea's a work in progress, bear with me and my explanations for now. ^_^

Oh, but ironically I've actually switched to ungraded cards in the SRS for those other subjects I mentioned, I don't even add fill-in-the-blanks, I just copy/paste pieces of text and read them, then pass them. If a passage comes up and it 'feels' too unfamiliar and doesn't summon conceptual associations, then I fail it or grade it hard. It's odd but works.

nest0r Wrote:So I'm curious, how do you folks use 'cloze deletion' in Anki (or outside it)? I am thinking of using it for light novels somehow, perhaps selecting a swath of sentences per volume and SRSing them (with Question side having sentence in cloze deletion form, word on the Answer side), or perhaps SRSing only the extracted vocabulary and then reading the novel normally (but with the SRSed words removed).

nest0r Wrote:Thanks for the replies. Seems there's a lot of possibilities. I've tossed a few ideas around over the past cpl years but haven't actually used them .;p The way I've used cloze deletion with other subjects is while I'm reading, I copy/paste a passage into the SRS, cloze delete a few salient points so that the single words are on the Answer side... the goal was only partially to remember particular 'word-concepts', mostly I was just trying to anchor the ideas to one another, because the side-effect of that process was almost a rote memorization of the word when I read the swath of text I took it from.

So then when I was looking at how useful it was to offload goals for sentences (ie the listening/parsing in a video clip deck, the vocabulary in a separate deck), I wanted something similar to practicing reading--except for reading the emphasis wouldn't be on listening and parsing at someone else's pace, it would be to try and get used to reading multi-sentence areas of text without stumbling, getting used to a less strict internal process than I would have with the more intensive recognition cards I did with smart.fm/KO2001 starting out.

In that sense, I wouldn't be doing cloze deletion for reading to memorize vocabulary so much as to get used to the process of associations words with one another quickly, making inferences on the fly, etc., and the 'cheating' aspect of memorizing stuff by rote and having those 'hints' would actually ensure a certain amount of stability, as long as I maintain a smallish pool of words and paragraphs. Anyway, the idea's a work in progress, bear with me and my explanations for now. ^_^

Oh, but ironically I've actually switched to ungraded cards in the SRS for those other subjects I mentioned, I don't even add fill-in-the-blanks, I just copy/paste pieces of text and read them, then pass them. If a passage comes up and it 'feels' too unfamiliar and doesn't summon conceptual associations, then I fail it or grade it hard. It's odd but works.

Idea's still a work in progress (i.e. I forgot about it till someone mentioned 'MCCD' in AJATT), but another thing: dynamic/multiple cloze delete collocations

BTW is this Dictionary of Japanese Collocations any good? http://www.amazon.com/Dictionary-Japanes...4770023006

Also, Irixmark mentioned this, Common Japanese Collocations: http://www.japantoday.com/category/book-...llocations - Any good?

So, how about a script that generates collocations from ebooks and other digital materials?

I have the second book. (I had a coupon at Borders.) Not every phrase gets a sentence, but there are a lot of collocations in there. Also, collocations are broken into 6 main themes: Home; Daily Life; People; School and Work; Travel; Entertainment. Then each theme is subdivided further into what I'd call "scenes" like "Body & Hygiene"; "Makeup"; etc.

So I'm not really sure how useful it is as a reference book. There's no index, so you have to figure out what you want do, where you want to do it, figure out the scene, then hope they included it.

It feels like going through someone's attic.

Some Umbrella examples:

Umbrella: 傘　かさ
傘をさす　open an umbrella
彼は大雨の中を傘もささずに歩いていた。　He was walking in the heavy rain without an umbrella.
傘を開く[閉じる/たたむ]　open [close] an umbrella.
傘をすぼめる　close an umbrella halfway; have an umbrella half-closed.
傘を持って行く　bring an umbrella
天気予報では夕方から雨らしいから、傘を持って行きなさい。　The weather report says it will rain this evening, so take your umbrella with you.
... (skipping some more obvious ones)...
傘をさしかける　share one's umbrella
可愛い女の子が傘をさしかけてくれた。　A pretty girl shared her umbrella with me.
傘に入れてもらう　come under someone else's umbrella
彼女は傘を持っていなかったので、駅まで傘に入れてあげた。　She didn't have her umbrella with her, so we shared mine to the station.

Hopefully that will give you an idea.

Is it just me, or in this day and age, shouldn't Japanese learning texts come in OCR'd ebook form. I would buy those books but just can't be bothered with dead tree materials.

No, it's not you. I feel the same way. I have a stack of books I want someone to scan and OCR for me, but I don't want them disassembled in the process. :\

I especially want to take some of my grammar problem set books and turn them into cloze cards for "fun"... and that's going to either require a) a lot of typing of b) one of the Japanese-branded OCR programs, which won't die when it runs into underlined text or a _____. (Unlike IRIS.)

nest0r Wrote:So, how about a script that generates collocations from ebooks and other digital materials?

Those exist for text - some free, some paid. The guys that are into corpus analysis use them. Some let you set how many words before and after to consider and it spits out all kinds of fancy frequency stats, etc. You can also enter a number of text files and then generate a list of sentences in which word A appears within so many words of word B, etc. When you click one the sentences, it open up the source text at that point.

I played around with Laurence Anthony's free concordance and word profiler software ages ago. (He's at Waseda U) Not sure if it has been updated or how it compares to what else is out there. The words profiler is based on vocab guru Paul Nation's Range program, but for Japanese. I'll see if I can find the website that lists a whole slew of them.

As for dictionaries, there's also:
研究社 新編 英和活用大辞典 (EPWING CD-ROM 版)
The Kenkyusha Dictionary of English Collocations on CD-ROM