Suggested Courses Feature

It would be nice to have a feature that suggested courses for you to take based on what others are taking that have finished or are doing the courses you have. For example, if you just finished course "animals" and most people that have done that course do "jobs" the suggest feature would suggest you do "jobs" next.

Or it might suggest courses based on the fact that another course shares some of the words you already know.

I know this might involve some heavy affinity analysis, but it would be pretty cool.

Posted by jlptbootcamp 5/6/12, last update 5/7/12 (2 years ago)
  • I think rather than a topic suggestion (which would make it necessary to tag courses), markings on frequency in a language should be used.

    Let's say you just learned some bird names, some transportation nouns and some medical terms, along with the vocab of the two textbooks you use, you could "fill the gaps" with a deck generated specifically for you by estimating your vocab size and use the words not in any of your lists, within this range.

    With a frequency list, the vocab database and a decent script, words could be flagged in a heart beat. You can generate lists of any size picking unused words from that pool for the optimal way to broaden your understanding - and deepening it over time, as you progress.

    If a user could also flag words as "I know this and see it often enough, so I don't need to see it here on memrise", you can then go through vocabulary in a much more meaningful way.

    Why would I want to "learn" and repeat "hello" over and over again? I don't even have to see it every 3 months, because I actually do use the language.

    That's what I'd do.

    Posted by ikenaiAndi 5/6/12 (2 years ago)
  • I was just trying to think of something a little easier to implement as the first generation of the tool. Although, you could create frequency lists by looking at how many people have learned a particular word. The more people that have learned it, the more frequent it is assumed to be. Then you could you use that to generate the lists you are talking about.

    Do you know of any good frequency lists for Japanese? I haven't been able to find any. I know you can pretty easily find English ones though.

    Posted by jlptbootcamp 5/6/12 (2 years ago)
  • Just load up ANKI, there you go. Core 1.000, 2.000, 6.000, 10.000 ... there should be a 20.000 somewhere on the net, too. The only problem with these lists is the material they're based on (mostly newspapers), so what you learn will be useful, but probably not the same people use on the street. That's why I plan to build up the Jpod list after I finished the C6K.. to fill my gaps :)

    Hmm.. Not sure which is easier to implement TBH.

    Posted by ikenaiAndi 5/7/12 (2 years ago)
  • I was talking about something a little more granular, I know about the Kore lists, but in order for the suggested courses to work well, it would have to be something a little more exact, don't you think? Not everybody is a big studier like us. They would want something a little more bite-sized, like what are the next 100 words that are most common for example.

    Again, I think this could be pulled from the data that is available on memrise. The more people studying it should be equal to how useful the word is and thus how often it is used (at least by the people on memrise).

    Posted by jlptbootcamp 5/7/12 (2 years ago)
  • Hmm.. you have a point there. Then again, this only works if many people learned a lot of words. I tested a few lists and abandoned them again shortly after, but some people have gardens with 50 lists, each only partially learned.

    Now, for a topic like "Japanisch", with roughly 30 users and 2 sets, such suggestions are just not feasible - English rules this world with an iron fist, and even the default display settings are biased towards this, here on memrise. If you are to search for your TL through your mother tongue, you'll have to figure out where the site hides them ;)

    I do like the idea of 100 word minisets generated by memrise. There must be ways to somehow deal with the problems related to that.

    Posted by ikenaiAndi 5/7/12 (2 years ago)
  • WE actually were just discussing how to implement this kind of thinking based on frequency data, and were running up exactly this issue. For French, Spanish, German, English and Chinese we can do ok on the frequency data (though the data is always a bit messy, it is good enough). For other languages is gets harder. If we could get data for a good number of languages, then it would make it easier to spend the time implementing it.

    So any suggestions would be very welcome!

    Best wishes

    BEn

    Posted by BenWhately Staff   5/7/12 (2 years ago)
  • I know there are "official" frequency lists for Japanese (as official as that can possibly be), based on newspaper articles.

    The University of Leeds has put up quite some frequency lists for various languages (including Japanese, also a Kanji frequency list) in sorted order here:

    http://corpus.leeds.ac.uk/list.html

    Posted by ikenaiAndi 5/7/12 (2 years ago)
  • Just found out they used the internet as the corpus for determining frequency - that's great news!

    Posted by ikenaiAndi 5/7/12 (2 years ago)