I am a course contributor for the four-part course, “8,000+ Most Common Swedish Words”, originally created by a guy called “sehiralti”, who seems to have pretty much moved on, from what I can see (the number of points he has racked up occasionally moves upwards, so I know he uses memrise from time to time.
Anyway, this is my problem: when sehiralti originally created the course, he used a database called the Kelly list (“SweWac”) as his source and created four individual courses using this data. At some point, a memrise user on the forum pointed out that one of the entries in the most popular course (11.7K have signed up for it) was incorrect, namely that “att andra” should have been “att ändra”. This prompted me to finally get round to printing out the database at home so I could check for myself. I began the laborious task of comparing the entries in the course and those in the database and found that a number of changes had been made, some deliberate and some just typos (like the above one, that an “ä” had been seen as an “a”).
By this time (last year sometime), I was into my third year of learning Swedish and was in a position to judge what had happened in some cases, which was this: for a number of seemingly “duplicate” entries, “sehiralti” had replaced the duplicates with words from another database, many of which were numerals.
So I began to replace the incorrect entries with the correct ones from the original frequency database, “SweWac”.
As a result, we now have entries with inconsistent mems. To give you an example, one incorrect entry was “femton” (Swedish for fifteen), so there a bunch of mems featuring the number fifteen at that point, which are just clogging up the course and don’t serve any useful purpose any more.
I have tried “flagging” them, but that is a complete waste of time and has never resulted in any changes as far as I can tell: the same old irrelevant mems still come up 
The other problem with some of the mems is that, due to the single-word definitions which were originally offered for the most part (an understandable solution, I hasten to add, given the size of the project that sehiralti tackled), some learners misunderstood which meaning was required.
A case in point would be this one: the word “en kull” had been translated as “a litter” and a number of learners had understood it in the sense of “trash, garbage” etc and made their mems accordingly, choosing pictures of trash cans and so on. Later, I discovered that the word refers to a litter of baby animals 
Here’s another kind of unnecessary mem that I see all the time: the person in question merely duplicates the definition of the word - and nothing else - and makes a “mem”. Obviously, these are really of no use to anyone and the people who make them have no clue what mems are supposed to be about. These could also be removed.
Another kind of unnecessary mem I can think of are the ones that came about when memrise was only available as a web version and images were offered. As a result, for some words, there are multiple versions of the same image, sometimes with text, sometimes without.
Finally, there are all the mems that contain misspellings and incorrect grammar. Those could also be removed or improved if we course creators had the permission to do so.
It would be great to have permission granted to course creators to remove mems so that these kinds of unnecessary and misleading mems could be removed.
