I'm taking my time now to learn Ruby and Python.
I'd like some ideas for side projects.

What tools do you wish existed that would make a real impact in
Japanese learning experience? Anything unrelated to Japanese learning
is OK too.

Thanks.

I think that depends on your current level with the language. At my current level, just watching dramas / movies with subs and using a tool like yomichan to import directly into Anki whatever I don't know is exactly what I need (as well as using Anki for drilling grammar, sentences, etc.)

Perhaps a tool like yomichan/rikaichan that would also search Weblio. I always find there what rikaichan fails to report.

Language games that involve drilling vocab, readings etc. I'm thinking of something along the lines of an open source version of iKnow's Brainspeed, with anki-like configurability (sort of like a cram mode on crack).

I would want a game that would allow you to choose the cards, add others later... some sort of customization feature integrated into the game play. I personally like the Anki-Emperor/Warrior game of "Empire Building," but I would imagine there are countless possibilities.

Some project ideas or things on my todo list:

- A better word frequency list. I currently use a list based on anime and drama subs, cb's novel frequency list, and internet-jp-forms.
- A TSV file and Anki deck for vocabulary to study after Core 6000, with audio files from WWWJDIC.
- A shell SRS for learning RTK keywords. If for example it showed 字, you'd have to type "character" for a correct answer. It could store the repetition times and grades in a text file.
- A Safari extension like Characterizer.
- A shell script that uses ImageMagick and ffmpeg to create videos for reviewing vocabulary.
- A DIY KAKASI based on MeCab.

(Edit: I have finished these and posted them on my website.)

- A script for estimating the difficulty of sentences based on word frequency.
- A script that modifies Japanese subtitles to add English translations for uncommon words.
- A list of dictionary definitions for uncommon English words used as first translations in EDICT.

lauri_ranta Wrote:- A script for estimating the difficulty of sentences based on word frequency

Kinda related, but based on JLPT level, not frequency: http://language.tiu.ac.jp/tools_e.html

Probably something entirely automated, involving an IV drip; controlled release of dopamine; direct sensory stimulation; a comfy chair and the reassuring voice of Morgan Freeman in 大阪弁.

anritsi Wrote:- something like PotPlayer, but where JP subs copy to the clipboard automatically

Solved one of my own wishlist items through a forum search, haha: Capturing Japanese subtitles with AGTH, since KMPlayer seems to work fine with ITH.

anritsi Wrote:but I'm mostly eagerly awaiting overture2112's promised vn2srs (visual novels 2 srs)

@overture2112 lol! I was definitely not expecting anything right away. I'm actually still going through Core6k & just leisurely reading everything else atm (ie, I look up defs if convenient but don't do anything further). Maybe after 2 months or so~ :)

anritsi Wrote:- a modded rikaiSama that could tell me when I should know a word (ie, it's already a seen one in my Anki deck)

I haven't messed around with firefox development much (and technically have been using mostly chrome lately) but I was looking into doing a few things for browser:

1) Select a word and hit a key to jump to sentences that contain that word in your anki card browser. The idea being to then pick one of the sentences and queue it for immediate learning (via the new morph man feature).

2) Color all the morphemes in a webpage according to difficulty (ie. a more general application of the new morphHighlight feature I'm working on).

I haven't thought about these too much but maybe rikaisama is the place to add them.

anritsi Wrote:but I'm mostly eagerly awaiting overture2112's promised vn2srs (visual novels 2 srs), *excitement~~* \o/

I can't figure out how to package it into a nice solution. It's not particularly difficult to understand how to set it up, but it's rather involved as there's many pieces that have to work together and certain parts don't have a general solution (e.g. how to provide an english translation for lines) so it'll probably be released as something more like a kit. I'll try to post about it later today and release something this weekend if the MorphMan update doesn't take much longer.

overture2112 Wrote:1) Select a word and hit a key to jump to sentences that contain that word in your anki card browser. The idea being to then pick one of the sentences and queue it for immediate learning (via the new morph man feature).

Oh yes, that would be very very nice. In combination with an up-to-date Rikaisama that works on Mac that would be absolutely fabulous.

I thought about this and realized there's only one thing I can think of that would make a big difference to how I study: something that displays Japanese words as I speak them. I'm sure there's plenty of speech recognition software out there, but I have no idea what works with multiple languages and most importantly what works well.

I imagine a tool that lets me speak into it and then displays Japanese words with the correct kanji (or in kana if I prefer: there should be a simple way of selecting between various options like standard IMEs have), and also can do the same for English. There would be a dictionary feature which will display the definition(s) of a highlighted word from all installed dictionaries (for example, it would take epwing format files or something) and that definition, together with example sentences if they exist such as in Kenkyuusha, would be able to easily be copied and pasted anywhere.

This tool would have to work with Anki as I'd be using it to quickly create my cards; at the moment card creation is one of my biggest problems as my cards include among other things the kanji word/expression if it needs to be learnt, the kana reading for that, the English translation (or Japanese definition if I feel it's more useful), and example sentences showing the usage of that word/expression. So it takes me ages to make a card but I can't think of a better way that allows me to learn example usage at the same time. I do extensive reading and listening but don't feel confident I would learn well enough from that. If only there was a truly good Japanese speech recognition tool. Maybe there is a way of doing all these things but I just don't know about it, I haven't researched everything about it yet.

I'd most like an extra difficulty button for Anki that'd, say, let me manually enter the amount of time till I next saw a card. Currently doing RTK for the second time, and a lot is still in there somewhere really, so after seeing some cards a couple times I'm confident enough I don't actually need to see it again for a while. It's driving me absolutely nuts how often cards I know well come up before finally reaching longer intervals, even if I press Easy/Very Easy every single time. I can use the reschedule function, but that takes more time...and the first time I used it I succeed in completely messing up the review intervals for all my cards, which makes me rather reluctant to touch it again.

I'd call it the 'So easy I just want it to go away and leave me alone' button. Or maybe not. XD

Ampharos64 Wrote:I'd call it the 'So easy I just want it to go away and leave me alone' button.

Hit "@" (suspend note)

If you're going through an old deck and failing cards that used to be mature, there's a setting for letting failed mature cards retain some percentage of their old interval. It defaults to 0%, but perhaps in your case make it 10-20% or something? That said, it should only take like 6 (maybe less) reviews to get your interval to >2yrs if you hit very easy every time.

Heh, I don't really want to suspend, them, though, I want to see them again...in a few months or more. If I suspend them I think I'll forget to unsuspend them, and it would involve a lot more manual management since the interval I'd want to set would vary - many are Easy, but some are five months Easy, and some the only reason I'm even really bothering adding at all it is for the keywords/primitive meaning, to support my reminding myself of other, harder kanji. It's a new deck, but as I'm going through RTK again, it's turning out that among the kanji I've totally forgotten, there are many more I actually only need a couple reminders of.

Could try being patient, I guess, yup, it's just when there's so many kanji even six reviews (may be more than that, it feels like it at least) really piles up, and it's making the whole process very boring.

karategirl09 Wrote:I'm sure there's plenty of speech recognition software out there, but I have no idea what works with multiple languages and most importantly what works well.

I looked into this a while ago, and just about every freely available solution for Japanese is terrible. Even text-to-speech is terrible. There's no chance for multiple languages using the same system when the languages differ as much as English and Japanese. And forget about cross-platform.

It's unfortunate.

Ampharos64 Wrote:let me manually enter the amount of time till I next saw a card

You could mark the card (* button), then hit the easy button. Then once a week or so, you can go to the browser and manually reschedule all marked cards in bulk.

Or change the easy bonus and avoid using the easy button until you get one of these cards. There's also this magical "delete" button.

When I was doing Core 6K with JP word + sentence on front and English word + sentence on back, I thought it would be nice if the sentence could be randomly selected. Otherwise, the sentence would often give away the word, so felt like I wasn't really learning the word. That wouldn't be an issue if the sentence was random. I'm not sure if something like this can be accomplished in Anki or an Anki plugin.

Ampharos64 Wrote:Heh, I don't really want to suspend, them, though, I want to see them again...in a few months or more. If I suspend them I think I'll forget to unsuspend them, and it would involve a lot more manual management since the interval I'd want to set would vary - many are Easy, but some are five months Easy, and some the only reason I'm even really bothering adding at all it is for the keywords/primitive meaning, to support my reminding myself of other, harder kanji. It's a new deck, but as I'm going through RTK again, it's turning out that among the kanji I've totally forgotten, there are many more I actually only need a couple reminders of.

Could try being patient, I guess, yup, it's just when there's so many kanji even six reviews (may be more than that, it feels like it at least) really piles up, and it's making the whole process very boring.

I feel like if the card is that easy, then you can hit "Very Easy" in half a second and move on. If it takes more time than that to figure out, then it's not actually "Very Easy", and if it does take half a second, then the time being wasted is negligible. Maybe that's just me.

headphone_child Wrote:When I was doing Core 6K with JP word + sentence on front and English word + sentence on back, I thought it would be nice if the sentence could be randomly selected. Otherwise, the sentence would often give away the word, so felt like I wasn't really learning the word. That wouldn't be an issue if the sentence was random. I'm not sure if something like this can be accomplished in Anki or an Anki plugin.

I don't get what your saying.

"Make the sentence randomly selected"? Are you saying, pick a random sentence that has the word you are trying to learn? Where are you getting the sentences? How do you even know the word matches the meaning for that specific situation?

If what you are saying is that "You know the word meaning because you memorized the sentence." Then I would say, just keep at it. At some point you are going to hit 2000-3000 sentences and suddenly, remembering a word based on the sentence would be difficult.

If you are saying you used the stuff in the sentence to figure out the word, then...why is that problem? Thats context reading, pat your self on the back, thats a skill you want to try and develop.

vix86 Wrote:

headphone_child Wrote:When I was doing Core 6K with JP word + sentence on front and English word + sentence on back, I thought it would be nice if the sentence could be randomly selected. Otherwise, the sentence would often give away the word, so felt like I wasn't really learning the word. That wouldn't be an issue if the sentence was random. I'm not sure if something like this can be accomplished in Anki or an Anki plugin.

I don't get what your saying.

"Make the sentence randomly selected"? Are you saying, pick a random sentence that has the word you are trying to learn? Where are you getting the sentences? How do you even know the word matches the meaning for that specific situation?

If what you are saying is that "You know the word meaning because you memorized the sentence." Then I would say, just keep at it. At some point you are going to hit 2000-3000 sentences and suddenly, remembering a word based on the sentence would be difficult.

If you are saying you used the stuff in the sentence to figure out the word, then...why is that problem? Thats context reading, pat your self on the back, thats a skill you want to try and develop.

Yeah, I'm talking about getting a random sentence containing the word. The sentences can be mined from anywhere, like alc.co.jp or jisho.org. Of course the sentences should be downloaded and included in a deck somehow rather than queried online each time. Knowing whether the sentence has the correct meaning of the word is indeed a technical challenge.

Here's an example of why I thought this would be useful:

すらすら -- その小学生は難しい本をすらすら読んだの。

I could see すらすら by itself and not remember what it means, but if I look at the sentence, I could remember "there was a gradeschool kid that could read tough books easily". So I got this without remembering すらすら at all. That wasn't using context to figure the sentence out. That was me remembering that sentence instead of the meaning of the word. If this sentence was different each time, and I used context to figure out the meaning each time, I'd be fine with that, because that develops that desirable skill as you mentioned. But that's not necessarily what happens when you get the same sentence each time. That's why I brought up the topic of random sentences.

Perhaps removing the English sentences would help you? If each sentence have one new word, you shouldn't need the full sentence translation to understand it.

headphone_child Wrote:That wasn't using context to figure the sentence out. That was me remembering that sentence instead of the meaning of the word. If this sentence was different each time, and I used context to figure out the meaning each time, I'd be fine with that, because that develops that desirable skill as you mentioned. But that's not necessarily what happens when you get the same sentence each time. That's why I brought up the topic of random sentences.

I get what your saying because its a worry I had. I ended up changing my core6k deck though to try and combat it. I bolded and underlined the keyword in the sentence. It stood out better and I eventually trained my self to hone in on that first. This helped me stop memorizing the sentence and put the word first. That said, if you can see the word in the wild and remember "Oh that was with the grade school sentence and it meant X" then what does it matter.

That being said, that technical challenge is the main thing making this very unlikely. A lot of stuff probably won't line up semantically when choosing at random.

EDIT: Theres also an issue that finding an n+1 sentence might be hard.

vix86 Wrote:

headphone_child Wrote:...That was me remembering that sentence instead of the meaning of the word. If this sentence was different each time...

...I bolded and underlined the keyword in the sentence. It stood out better and I eventually trained my self to hone in on that first. This helped me stop memorizing the sentence and put the word first.
...
That being said, that technical challenge is the main thing making this very unlikely. A lot of stuff probably won't line up semantically when choosing at random.

I suggest you try out the new Morph Man v3.3. There are three relevant features that you might be interested in.

1) As per vix86's underlining suggestion, it can automatically highlight the unknown word in a sentence. In fact, you can arbitrarily style morphemes according to maturity level in category (eg. seen but not known, known but not mature, etc) or the interval number itself (eg. 4, 7, 19.3 days, etc) with CSS.

2) Create a filtered deck for "tag:comprehension -is:review", which will get you m+0 sentences you're not actively studying and burn through some to practice recognizing words and grammar points in different contexts.

3) While reviewing a sentence with a "focus morpheme" (ie. the single unknown in a k+1 sentence), hit "L" and test yourself against the "alternative" sentences, which are other k+1 sentences that also contain the particular vocab you're focusing on.


What you said you wanted is essentially a more automated #3, which can be accomplished by something like:

Code:

from anki import sched
import random

def my_getRevCard( self, _old ):
    # 1) get next card according to anki's normal method
    card = _old( self )
    # 2) get it's focus morpheme if it has one
    try:
        focus = card.note()[ 'focusMorph' ]
    except KeyError:
        focus = None
    # 3) if it has one, we look for alternatives
    if focus:
        searchQuery = "focusMorph:%s -is:review" % focus
        alternativeCards = mw.col.findCards( searchQuery )
        # 4) if it has alternatives, pick one randomly to replace it
        if alternativeCards:
            card = random.choice( alternativeCards )
    return card

sched.Scheduler._getRevCard = wrap( sched.Scheduler._getRevCard, my_getRevCard, 'around' )

Note the above isn't a full solution, you still need to modify what happens when you answer a card so that it modifies the scheduling of the original card instead of the one you replaced it with, but that shouldn't be too difficult.

A modded subs2srs that would gloss the sentences with J-J definitions from Yahoo辞書 instead of making me do that manually; making it just as time consuming as normal sentence mining.