Oh. My. God. I've been away from RvTK completely for the last couple weeks since I'm studying for the kanken Feb 3rd (grade schoolers know a f*ck lot of words!!) so I hadn't read any of your updates. Trinity looks AMAZING! It is exactly what I have imagined for a review system. Anki is close, but it scatters learning goals that are intertwined (sentences, vocabulary, readings), and the entry process, while flexible, is often cumbersome.

I think Trinity remembering kanji/vocab you already know and cross-linking them wherever they appear is huge. It's what makes computer review better. Information can simultaneously exist in multiple "places" for mulitple purposes and is not bound to a static location. I shouldn't have to enter 小学校 4 times, once for the word, and 3 times for each kanji, and then have to enter to every time I may use it in a different example sentence.

The mapping feature for the on-yomi is genius. You obviously understand the benefit of incorporating multiple learning methods to support retention.

Will the sentence entry feature automatically recognize vocabulary? If they have to be selected, there is at least a simplified process of clicking and picking, right?

So for reviews, does it just quiz you as an SRS for each entry title? Kanji and vocab are obvious. Will sentences work the same? You will be presented a sentence/"word list" and you fail or pass it if you successfully read it?

I know it's a longshot, but is there any way that Trinity will be available offline? Could there be some sort of database syncing, at least for reviews, that allows you review in a browser offline but have the site recognize your progress afterwards? (Entry could be kept online since that involves so many other databases, but would be megasweet if it was offline too!)

Btw, could this be done by, say, yesterday? I'd love to use it for my kanken reviews.

meolox wrote:

Excitement rising, really hope I finish by the time this comes online.

Thanks! Well if you haven't finished it can be helpful for RtK1 too. I don't bother clearing my red stack anymore, I just clear the characters when they come up in a word I'm adding in Trinity. This has the advantage that the word helps remember the character. Sometimes there's a blurry line between the compound mnemonic that naturally comes out of the keyword combinations, and the story for the kanji itself, but that's a minor problem.

dingomick wrote:

I shouldn't have to enter 小学校 4 times, once for the word, and 3 times for each kanji, and then have to enter to every time I may use it in a different example sentence.

Thank you, looks like you've figured it all pretty much   To be precise, you will have to enter a word for every sentence it appears in, if you find this association to be useful. If the word is already familiar, you don't have to attach it to the sentence. If you don't remember for sure what the word means, chances are you didn't study it yet.

Trinity does not currently parse the sentences. It would be difficult, but possible to use a specialized program to parse the sentences and extract the related vocabulary. But there are also advantages to doing it manually. One is that the user can decide what words to attach and limit the attached vocabulary to those that are new or challenging at that point. I think having an exhaustive list with each sentence , listing all the particles, and common adjectives or verbs that may be present would make the related vocabulary less effective, particularly when you review sentences, as it will show the linked vocabulary when you flip the card.

Another reason it's not automatic is that unless you're a beginner, you don't want to create hundreds of flashcards for words you know already. So you can add only those words that you would want to review.

Will the sentence entry feature automatically recognize vocabulary? If they have to be selected, there is at least a simplified process of clicking and picking, right?

It's easy to enter words... if you know already how to read them, that's the trick. Otherwise you might have to do some copy/paste from the sentence to the search box, to lookup the word with kanji. You can use cursor keys to browse the lookup results and the enter key to add a word, or click the words directly.

Will sentences work the same? You will be presented a sentence/"word list" and you fail or pass it if you successfully read it?

Sentences have their own flashcards. So you have two types of flashcards in Trinity : vocab and sentences. How you rate your answers for sentences is up to you. You may want to pass if you remembered the reading alone, or to pass if you remembered the whole meaning. If you do use sentences a lot, it would be better to pass vocab even if you remember only the reading (especially new exemplary words added while learning onyomi groups), but pass sentences only when you remember the full meaning.

I know it's a longshot, but is there any way that Trinity will be available offline? Could there be some sort of database syncing, at least for reviews, that allows you review in a browser offline but have the site recognize your progress afterwards?

It's possible with Google Gears for example. I didn't feel it was worth the effort to do this for RtK1. Well, who knows maybe it's not so hard to implement, I have yet to look at it in details. But for Trinity this is something I really want to do, though I wouldn't wait for it.

synewave wrote:

Will we be able to import lists into Trinity?

Sure, but it's not in there yet. For the Trinity beta, which could last 2-3 months I want to see how people use it from the grounds up. I 'm not sure I will add import during that time but who knows. Importing CSV's should be fairly easy.

Hi Chadoro_K

You're right about the RtK3 kanji. In fact some onyomi groups become substantially larger, the largest one, コウ contains ~66 kanji within RtK1 but increases to ~109 with RtK3. At the same time, the total number of groups only increase by ~15.

Chadoro_K wrote:

Perhaps a "Include RTK3" checkbox could be created. This way those who have done both RTK 1 & 3 can have the entire ON subset displayed while those who have only studied RTK1 would only see the RTK1 subset of ON readings.

There will be support for RtK3 as I have been thinking about letting the user choose different "targets", such as the general-use kanji. While I have to think of the implementation of those ahead (for setting up the database etc.) I can't prioritize a RtK3 target at the moment. It won't be in the beta of Trinity.

I also need to think about a way for users to add any kanji with the correct reading into a kanji chain, if someones wants to incorporate the alternative readings; that is one character in multiple onyomi groups.

I think it's efficient to do a long chain at once, but it will be difficult to study it all at once for groups like コウ, especially adding an exemplary compound for each. Once you have a chain and a story or spatial relationships between the characters, I don't see any reason why you couldn't add to it. When you'll "extend" a chain later you'll have to be careful to connect the characters to the exisiting plot.

The chain editor with its visual interface and spatial relationships can make this easier because if you place a character on the "map", you can see immediately closeby characters and can adapt your story or mental image to create a relationship between them.

So any news on the Trinity front? ^_^

I love the idea of sharing graded (or un-) texts!  One thing I've recently been enjoying on this site is the Reading practice - and I was just thinking that it would be great to be able to save all the texts I've inserted so I could go back to them and use them for studying.
Thanks!

Is there anything that could use some more coding hands, Fabrice?

ファブリス、

You might want to get in touch with resolve, since Anki successfully uses Kakasi to determine possible readings for entered sentences.

If you download the Anki source and take a look at the file libanki/anki/support/japanese.py you'll find code which invokes Kakasi for this purpose.

ファブリス wrote:

Hi Chadokoro_K

There will be support for RtK3 as I have been thinking about letting the user choose different "targets", such as the general-use kanji. While I have to think of the implementation of those ahead (for setting up the database etc.) I can't prioritize a RtK3 target at the moment. It won't be in the beta of Trinity.

I also need to think about a way for users to add any kanji with the correct reading into a kanji chain, if someones wants to incorporate the alternative readings; that is one character in multiple onyomi groups.

ファブリス,
A belated thank you for this reply.

It is great that Trinity will (eventually) support RtK3 and other targets (such as general-use kanji).

And it's also great that Trinity will allow users to add any kanji with the correct reading into a kanji chain. (Yeah!)

Thanks for all of your hard work on this!

I realize that you have a lot to do to even get the beta online but I think I speak for most everyone when I say,

首を長くして待っています。 

Last edited by Chadokoro_K (2008 January 28, 2:50 am)

Just read the Trinity news for the first time and it sounds awesome. I lost my sentences for Supermemo due to a defect HDD. Time to try something new. The concept sounds great, a system that includes most areas you need for your Japanese studies (Kanji, vocabulary, sentences). I can see myself spending hours per day on your website once Trinity is implemented (currently trying to add all the kanjis into my boxes ;D).

頑張って。

Hi Fabrice,

I use ChaSen with J-Gloss and find it "adequate" but, of course, not perfect. It is one of the very few free parsers available that will actually parse words written in kana (if not the only one).
A Windows version is also available with its own dictionary "ipadic" included.

Installation files may be downloaded from:
http://chasen.aist-nara.ac.jp/chasen/di … on.html.en

Cheers,

Tom

tomhogers: you might want to try to get jgloss to match up the kind-of-verb annotations ChaSen puts out with EDICT's v5r etc markup so you can avoid picking ones it can't be. Or at least have a special case for suru! :-)

Jim Breen talks about the Tanaka corpus here, and mentions some of the things he did with the ChaSen output:
http://www.csse.monash.edu.au/~jwb/papi … mples.html
The most obvious and easy one is that whereever ChaSen gives you two words you try looking them up in the dictionary as a single word, and if there's a match use that instead. (This counteracts its tendency to split compounds down further than is ideal.)

It ought not to be too hard either to just spot where it outputs lines for masu-form, past tense and the other obvious ones, and glue the relevant syllables back onto the preceding verb. (I had a quick scan of the docs but couldn't immediately spot anything to do this automatically.)

I note that you can tweak its output format with -F (help in -Fh), which is probably a good idea for this kind of automated usage.

ファブリス wrote:

It's possible with Google Gears for example. I didn't feel it was worth the effort to do this for RtK1.

My biggest gripe with RtK1 is that I need an internet connection, hence, cannot use it on the go while having lunch, in a cafe or something, meaning I can only really use it when I get back home, late, already dozing... and so kanji start to pile up :-/

Add my ballot to the "Yey for Google Gears ed.!" box. (^_-)

(sorry for the off-topic post, but since a ref. to Google Gears was made...)

Last edited by elmimmo (2008 November 17, 9:06 pm)