Minimum entropy is a confusing term

I’m admitting my ignorance here, but every time I want to relax the rules around unique characters in titles and post content, I get stuck on the descriptions of the body min entropy and title min entropy settings:

The minimum entropy (unique characters, non-english count for more) required for a post body.

There should be a way to put that into more human language. Looking at the code (discourse/lib/text_sentinel.rb at 63ca30ccb415d133c65aee6758b7f34d9705eea7 · discourse/discourse · GitHub), I’m not sure what to suggest. If the setting took an actual scale, say from 1 - 10, it would be easier to explain.

Feel free to close this if there are other topics dealing with the issue.

7 Likes

There is more human language in the alert of if a topic title isn’t allowed, but this isn’t clear in what the requirement for a title is.

Alert:

Didn’t know about that setting, that seems relatively clear that the number of different characters needed is what can be adjusted.

3 Likes

It’s tricky, because the setting is describing an algorithm, without supplying enough information to make it clear how the algorithm works. But yeah, it’s essentially saying “from 0 to (?), how much variation is required in titles and post content.”

3 Likes

I know what entropy means, but using those settings was just awful experience. No matter what values I tried it was constantly nagging. Setting itself is needed, perhaps, but because I’m a member of very small and very strange language-family, Finnish, and we use a lot vowels and those can be long, and same thing with consonants, I put zero on those.

Again. My forum is practically spam free and for normal grownup people, so I don’t even need it. I would guess situation isn’t as sunny everywhere. But I can see why it looks so confusing.

4 Likes

It should be a separate topic, but how to configure Discourse for adults is exactly what I’m working on at the moment.

5 Likes

Well, of course it depends if a forum is targeted to nordic, italian, south american (sorry for generalization), usa (big generalization here), academics, gaming, right, left… but I would like to say there is few universal rules:

  • too much decoration and gimmicks are just too much
  • Spam problems should solve only if there is need for that
  • Users must get those rights they need right away, so no playing with trustlevels
  • Kick bad apples right away out from forums, no warnings
  • Gamification is just another gimmick

And part of those is not using entropy settings.

4 Likes

This is a second alert have seen which can be good to ensure people are writing complete sentences:

However there is no way to answer this question, the system prohibits publishing of the post unless there is more entropy which isn’t explained to the writer.

With the minimum character limit for titles adding a few asterisks fulfills that limit but this additional entropy requirement could be a little more developed.

1 Like

My strategy is to have manual approval of the first few posts to filter out spam and other low quality content. All of the other limits like character count and entropy and all this other junk gets in the way of people engaging with the site so it’s turned off (or turned to an extremely tolerant threshold at least)

So far we have caught 100% of spam and no user has to suffer through a vague message about how their post needs more characters (which ironically probably promotes spam. Because to get around the message the simplest way is just to add character noise until you pass the threshold)

1 Like

The description “Minimum number of unique characters” sounds quite clear to me - i dont think you need a change here, especially given thats a really minor tuning. (however that NLS comment might be a bit confusing, does that mean it counts a national character as the number of UTF8 bytes it has?)

1 Like