[Site Feedback] The biggest learning inefficiency of using Memrise, database experts and Memrise staff do you think this solution will work?

I will take a guess that not a small portion of the people who leave Memrise do so because once you have a backlog it is difficult to bring it down again, because Memrise doesn’t work if you have a backlog. This makes this a problem worth solving.

I am curious why this wasn’t solved yet. Can you explain the difficulties to me, people who know how databases work and Memrise staff (@Joshua, @lien, @knarusk) ?

###The issue:

Any items you learn or review now gets added to the bottom of the review list. That makes it so that you will have long forgotten that item that has a review interval of 4 hours when you get to it next month.


Can you think of possible solutions that could work if you have lots of users and data?

One solution is this: currently Memrise uses column A to determine reviewing order, I propose adding a new database field (how long the item has been due for review / review interval) (column D):

You can see in the table that this prioritizes the items with a shorter reviewing interval, like it should.

2 Likes

First of all, I don’t know why everyone always says that new reviews get stacked behind the overdue reviews. That doesn’t seen to be my experience. I always have a backlog of several thousands of words to review, but I nonetheless seem to be learning and reviewing new words without any problems.

Second, I used to think this was a big problem, because i have like a kazillion items that I will never, ever get around to watering, and I wanted Memrise to create an algorithm to manage that backlog. I’ve recently changed my mind. If I haven’t learned an item after a year of reviews, then I probably just need to learn that particular item in a different way, through a different context, instead of reviewing it forever. Learn it, review it for a year or two, then just QUIT THE COURSE.

Memrise should be viewed as a temporary learning tool. Don’t expect it to handle years and years of reviews. There is no economic incentive for Memrise to create that functionality, and even if there was, we should be moving on to conversation groups, reading novels and newspapers, and watching movies. We didn’t spend 5 years reviewing our third grade spelling lists in our native languages, why are we upset that we don’t get that feature from Memrise?

2 Likes

Hi @lurajane,

As I am now working on my backlog that grew to over 10,000 items due to the behavior I describe, I can assure you that it does work like that. If you don’t notice it, it is perhaps because if you review all courses at once, Memrise rotates through them. So if you learn a word today, in a course where you have no backlog, Memrise does give it to you for review tomorrow. You can test it for yourself: go to a course where you have a big backlog and review or learn some words, then come back the next day and see if you see those words coming up for review then.

The rest of your post is on a different topic. To address it briefly, if I were to again decide to learn my native language or any European language, I would probably not bother with using any SRS at all and just start consuming native content and speaking and writing. I’m probably more extreme than anyone in that. I tried doing that with Chinese as well, but after a while I had to conclude that that didn’t work. I will again start doing that when I feel like it though.

1 Like

I fully unterstand you. I have the same problem.

However I have a script that allows me to skip memrises’ interval and use my real life time interval.

For example, if you learn a word today and only review it next week memrise will move it from 12 to 24 hours cycle.
This script will move it to the next cycle according with your real life time, in that case the word would go from 1 week to 2 weeks, because you were able to remember it after one week. Hope you understood the idea.

It is called catch-up-review and you can find it here if you want it:

1 Like

Hi @mikatu,

That is a different, but related, issue (I guess that would be the second biggest learning inefficiency). I am aware of that userscript and have used it in the past.