cb4960 Wrote:nest0r Wrote:cb4960 Wrote:Thanks for your assist!
By the way, did you catch my post here? http://forum.koohii.com/showthread.php?p...#pid140935
Don't want to overburden that gigantic brain of yours, but just something to keep in mind.
Just think, a counter for tracking our Firefox/Rikaisan encounters as a sort of reverse leech threshold, using the Realtime Import for adaptive Anki integration from there, so that's the non-SRS knowledge → SRS, but also the tracking of our Anki knowledge via MorphMan can be applied to Firefox, so there's the SRS → non-SRS! So close to the new flesh. ^_^ I mean oops, don't want to give away the master plan.
Ow, my head hurts. I guess the MorphMan thread is prerequisite reading.
With regard to rikaisan, I can certainly keep a history of looked up words and the number of times each word was looked up. Or were you suggesting something a bit more involved?
For the Firefox/Rikaisan → Anki part, that counter's what I meant. Would that allow for automatically adding cards to Anki using Realtime Import once a user set threshold is activated? Like, Rikaisan notes you've had to look up the same word X times and clearly incidental exposure isn't doing it for you, so it adds it to Anki. Hence the reverse leech threshold. (Personally I'd set it rather low but people will have their own goals/learning styles, and variables such as enumerated
here might come into play.)
For the MorphMan stuff, it uses Mecab to extract and create databases of morphemes/words for Anki which you can merge how you like, and you can use it to set and sort by i+N (how many unknown words per card, which you can store in their own field on the respective cards); ToasterMage made a script that apparently performs the i+N function outside Anki, e.g. on a text file.
So I was thinking this could be combined with JReadability. But I was also thinking this database thing is like what we talked about in the cbJisho thread for a level checker/reader using Anki knowledge, so I envision things like finding words in a light novel or online article using jReadability (?) and you use this information in various ways when reading using Firefox/Rikaisan and/or JNovel Formatter, so that not only does the counter/Realtime Import send cards to Anki, but Anki knowledge is also sent in the other direction.
With jReadability you could generate various statistics. Perhaps with JNF you can do that and/or format/sort/manipulate texts based on this information. Perhaps Rikaisan can tell you when you mouseover that you already know this word and other information about it (MorphMan also has a vocabrank function that I haven't really looked into, and there's ideas such as lexical profiling as popularized by Paul Nations' research into extensive reading and vocabulary, which is also used by that site/program Lextutor, apparently).
Edit 2: Or perhaps Aozora Remover, Gaiji Replacer, the Word Frequency tool, and jReadability can all be integrated? ^_^ With JNF? hehe...
Then there's the subs2srs stuff, where you can use known Anki information to filter new decks when creating them? Anyway, these ideas are just my usual idealizing for future developments.
Edit: Another thing that might've gotten lost in my previous comments in this thread is something for multiple layouts for the Realtime Import function in Rikaisan. Right now I have it set up for just my extemporaneous single word vocabulary deck, which I use purely for retrieving meanings at this point (word text + audio on the front, meaning on the back; context/writing/speaking/listening are all used as reinforcing/encoding information for that card type
Edit 3: Images too! That Image Download plugin for Anki is handy).
But we could possibly save and toggle to other layouts before adding, depending on what deck we have open? (Hopefully in the future we can tinker with when/where Realtime Import can actually add to Anki beyond having the specific deck open?)
Whew, so much for my timeout. ;p
Edit 4: More speculation: Hopefully overture (or somebody??) will be working on stuff related to maturity levels, e.g. that reminder plugin I've been desperate for, or automatic adding of cards to the databases or somesuch so you don't have to manually update them periodically. If that happens, we could possibly factor in the maturity stuff I mentioned before in that aforementioned cbJisho thread (60/30/10 type ratios), where you can see not just whether you ‘know’ information in outside files and use that information in various ways, but how well you know them (so that the i+N becomes sort of a gradient based on interval sizes/maturity levels/etc.; that's also leaving out other stuff MorphMan might do in the future regarding syntactic/aural complexity, but I suppose I rambled about that enough in the Mighty Morphology thread, re: lexical density, abd dependency parsing with KNP [by the way I think I/we broke the Langrid site, but there's an offline version of KNP linked in that thread, as well as a less cool web version]).
Edit 5: Ooh, and perhaps eventually with a counter you could start targeting information, e.g. parts of speech; or as I mentioned at the end of the MorphMan thread there's also the possibility of integrating that homophone/homograph plugin with Japanese Wordnet [which is also available offline].)
Edited: 2011-06-25, 3:07 pm
