Za-Chary wrote:

It makes users think more carefully about what to post before they actually try to post it. Thinking before you act/speak is generally a useful quality in life, and it is good to practice this on Scratch.

What if you didn’t know that there was a bad word? Pointing it out could teach you that it’s a bad word.

Za-Chary wrote:

It is harder to post bad/borderline content. If you knew exactly what was wrong with the comment, you could just change that exact part slightly and then quickly post a new comment which — if your first attempt was a “bad” comment — your next attempt would sometimes be a “mostly bad” comment. Why make small changes in this sense when you could just try to reword your comment completely? (I don't actually know if I said this point in the best way, but I know that from a moderator's perspective it makes sense.)

Well, we should really just fix the filter, as it already gives out false positives, and the community really doesn’t like the false positives.

Catzcute4 wrote:

Well, we should really just fix the filter, as it already gives out false positives, and the community really doesn’t like the false positives.

The filter is continually being fixed every single day. The filter is not perfect and never will be perfect. The Scratch Team does not intentionally put false positives into the filter, and in fact they would eliminate all false positives if it were possible to do so.

Sliverus wrote:

dertermenter wrote:

Yes, I just think my argument just has more merit to it since my example, whilst being rare, can happen at any time on Scratch - its an open window.

On the other hand, your example, can only happen on the rare occasion that the filter is down - and even then people who abuse this would already know the swear words they are typing - so an alert would not teach them any new very bad words, which is the main drawback of the topic.

First of all, who's to say they can't accidentally type one in, like you said, while the filter is down?

But more realistically, let's say “dog” is a bad word. When the filter is down, TONS of users abuse the lack of a filter. In fact, some users will even post words on their profile to test if the filter is down. There was a horrible situation like this that happened before while you were banned. Griffpatch's profile – or really, any popular community – was filled with inappropriate content.

You can bypass any filter whether it's down or not, just like how people bypass the filter for dinscord

(⌐⎔_⎔)//This is Jack, he's been fighting these evil kumquats since early 2023!


(((( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́))))  

dertermenter wrote:

snip

Your first point is a point. However, even the possibility of someone on Scratch being exposed to swear words, should never exist. One user being exposed to inappropriate content could be enough to get parents complaining, creating an article painting the Scratch Team in a negative light, and other parents/schools boycotting Scratch.

dertermenter wrote:

Your second point is a good point, however I think we need to realise how rare a filterbot outage is - there is a reason why the filter outage on 4th April 2022 has its own wiki page since its such a rare occurence (and also how important the filter is, but if this happened reguarly there wouldn't be articles about specific instances)

True, but I'm not referring to the filterbot outage for my second point. I mean in general, if the character limit in a Scratch comment is relatively small, and the user realistically knew what was in the copypasta, isn't it reasonable to assume that the user had experienced this swear word? And because any user that posts swear words has already seen the swear word, then it would be perfectly fine for the Scratch Team to call them out on it.

Za-Chary wrote:

I don't know why we're discussing what could happen when the filter is down when that has happened exactly 1 time in the past 16 years. To discuss a proposal for a feature in case the filter goes down seems like “doomsday speak.”

See my response to Dert's first quote in this post.

I thought of a compromise, by the way. What are your thoughts on this? If a user posts a word that is outright disallowed (like birth control things, for example), the filter will tell you what is wrong with the post. However, some topics are incredibly long (.e.g. stickies) and trying to edit them after massive revisions can be extremely frustrating. So if it is only subtly inappropriate (e.g. the cat example by @Za-Chary), the filter will tell you the sentence that contains the subtly inappropriate material.

Thoughts?

Sliverus wrote:

I thought of a compromise, by the way. What are your thoughts on this? If a user posts a word that is outright disallowed (like birth control things, for example), the filter will tell you what is wrong with the post. However, some topics are incredibly long (.e.g. stickies) and trying to edit them after massive revisions can be extremely frustrating. So if it is only subtly inappropriate (e.g. the cat example by @Za-Chary), the filter will tell you the sentence that contains the subtly inappropriate material.

Thoughts?

I might've missed something but why would you tell them which sentence it's in when at that point you might as well just tell the user what word it is?? It'd be a bigger hassle to find the word in a sentence. (don't even get me started on people who don't use punctuation)

(⌐⎔_⎔)//This is Jack, he's been fighting these evil kumquats since early 2023!


(((( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́))))  

doggy_boi1 wrote:

I might've missed something but why would you tell them which sentence it's in when at that point you might as well just tell the user what word it is?? It'd be a bigger hassle to find the word in a sentence.

It doesn't say what the bad word you said was, but it can help you find what to remove in order to be able to make the revision to your topic. Finding the word itself should usually be fine.

doggy_boi1 wrote:

(don't even get me started on people who don't use punctuation)

Why would someone not use punctuation in a very long post like a sticky? The only way I can imagine someone wouldn't use punctuation in a very long post like that is if they're spamming, and at that point it's obvious what's triggering the filter.

i support

Sliverus wrote:

However, even the possibility of someone on Scratch being exposed to swear words, should never exist. One user being exposed to inappropriate content could be enough to get parents complaining, creating an article painting the Scratch Team in a negative light, and other parents/schools boycotting Scratch.

In that case, it sounds like you don't support your own suggestion. If the filter tells you what you said wrong, there is a possibility that it would expose someone to a swear word.

Sliverus wrote:

I thought of a compromise, by the way. What are your thoughts on this? If a user posts a word that is outright disallowed (like birth control things, for example), the filter will tell you what is wrong with the post. However, some topics are incredibly long (.e.g. stickies) and trying to edit them after massive revisions can be extremely frustrating. So if it is only subtly inappropriate (e.g. the cat example by @Za-Chary), the filter will tell you the sentence that contains the subtly inappropriate material.

This assumes that it is easy to have the filter determine whether an offense is “outright disallowed” or “subtly inappropriate” — I don't think that would be easy, especially given the amount of false positives that exist. It also assumes that the Scratch Team can go back into the filter blacklist and determine specifically which words and phrases count as “outright disallowed” versus “subtly inappropriate,” which is definitely not easy.

Sliverus wrote:

It doesn't say what the bad word you said was, but it can help you find what to remove in order to be able to make the revision to your topic. Finding the word itself should usually be fine.

But why would they do that when you can just tell them what the word was. also see ZA's post above

Sliverus wrote:

doggy_boi1 wrote:

(don't even get me started on people who don't use punctuation)

Why would someone not use punctuation in a very long post like a sticky? The only way I can imagine someone wouldn't use punctuation in a very long post like that is if they're spamming, and at that point it's obvious what's triggering the filter.

because some people just don't. I wouldn't know why, ask them
edit: I messed up the quoting

(⌐⎔_⎔)//This is Jack, he's been fighting these evil kumquats since early 2023!


(((( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́))))  

Sliverus wrote:

I thought of a compromise, by the way. What are your thoughts on this? If a user posts a word that is outright disallowed (like birth control things, for example), the filter will tell you what is wrong with the post. However, some topics are incredibly long (.e.g. stickies) and trying to edit them after massive revisions can be extremely frustrating. So if it is only subtly inappropriate (e.g. the cat example by @Za-Chary), the filter will tell you the sentence that contains the subtly inappropriate material.

Whether the filter would know if something is “subtly inappropriate” or “outright disallowed”, as Za-Chary stated, would be near-impossible with the current system. That would require the filter to have context on what topics are considered inappropriate and to what degree they are

Za-Chary wrote:

In that case, it sounds like you don't support your own suggestion. If the filter tells you what you said wrong, there is a possibility that it would expose someone to a swear word.

Good point.

Za-Chary wrote:

This assumes that it is easy to have the filter determine whether an offense is “outright disallowed” or “subtly inappropriate” — I don't think that would be easy, especially given the amount of false positives that exist.

When did I say that implementing a new Scratch feature would be easy? Implementing a new feature isn't easy. But if the developers are willing to do the hard work, they could, perhaps, do this:

For the sake of the argument, imagine “top” is an offensive term.

If someone says, “You're a stupid little TOP!” – without spaces, without substitutions, whatsoever – the filter can show you the specific word you said wrong.

But if someone said, "I went to Patrick's house," where there's a space, then the filter will tell you that a word is wrong in a particular sentence. This would also happen with substitutions, but just to be safe, the specific word will not be named. It will show the sentence.

Za-Chary wrote:

It also assumes that the Scratch Team can go back into the filter blacklist and determine specifically which words and phrases count as “outright disallowed” versus “subtly inappropriate,” which is definitely not easy.

Again, it's not easy. But it doesn't matter whether it's easy. If they create this feature, it can open up new possibilities for features in the future.

doggy_boi1 wrote:

But why would they do that when you can just tell them what the word was.

I don't understand. Why would we do that?

doggy_boi1 wrote:

because some people just don't. I wouldn't know why, ask them

That was meant to be a rhetorical question. My point was that it doesn't happen. Can you find an example of an appropriate comment that was extremely long and was all one sentence?

ajskateboarder wrote:

Whether the filter would know if something is “subtly inappropriate” or “outright disallowed”, as Za-Chary stated, would be near-impossible with the current system. That would require the filter to have context on what topics are considered inappropriate and to what degree they are

Not necessarily. See my response to @Za-Chary.

Sliverus wrote:

When did I say that implementing a new Scratch feature would be easy? Implementing a new feature isn't easy. But if the developers are willing to do the hard work, they could, perhaps, do this:

There's a difference between “doing hard work” and “reaching for unattainable goals.” I fear this method would be best described via the latter. Also, with the method you've described: there are already so many false positives given by the filter, that there's not really a way that this method would fix things. There's no way to guarantee that your method wouldn't give false positives.

Za-Chary wrote:

There's a difference between “doing hard work” and “reaching for unattainable goals.” I fear this method would be best described via the latter.

How is this unattainable?

Za-Chary wrote:

Also, with the method you've described: there are already so many false positives given by the filter, that there's not really a way that this method would fix things. There's no way to guarantee that your method wouldn't give false positives.

I'm not trying to fix false positives. Maybe I should clarify. I want the forum filter to tell you what you said wrong.

Sliverus wrote:

- snip -

It is unattainable because, as had been said above, it is nearly impossible to have an automated filter detect context. I should clarify that here I am commenting only on your compromise where you want different filter detections to have different effects, not the main suggestion where you want the filter to show what went wrong. To detect the context in which a particular word is said sounds about as difficult as having a filterbot detect context in a comment. I recognize that your main suggestion is not asking to combat false positives, but in some sense your proposed compromise is.

Not to mention that the filter currently does not censor all bad words within all phrases in the first place, so the compromise wouldn't work as intended anyway. One can see this by the phrase “Occasionally I will say that sushi tastes great.”

Za-Chary wrote:

It is unattainable because, as had been said above, it is nearly impossible to have an automated filter detect context. I should clarify that here I am commenting only on your compromise where you want different filter detections to have different effects, not the main suggestion where you want the filter to show what went wrong. To detect the context in which a particular word is said sounds about as difficult as having a filterbot detect context in a comment.

I'm not asking that it detects context per se. But I'm not going to elaborate on this because I have a better idea instead.

What would say if the forum filter only told you the sentence that the inappropriate content, regardless of what the word is or the way it looks (e.g. substituting letters to bypass saying bad words)? The issue of context would be eliminated entirely.

Za-Chary wrote:

I recognize that your main suggestion is not asking to combat false positives, but in some sense your proposed compromise is.

Alright. I understand.

Za-Chary wrote:

Not to mention that the filter currently does not censor all bad words within all phrases in the first place, so the compromise wouldn't work as intended anyway. One can see this by the phrase “Occasionally I will say that sushi tastes great.”

I am not suggesting that any additional phrases should be added to the filter right now.

Sliverus wrote:

What would say if the forum filter only told you the sentence that the inappropriate content, regardless of what the word is or the way it looks (e.g. substituting letters to bypass saying bad words)? The issue of context would be eliminated entirely.

What counts as a sentence? If you mean “separated by a punctuation mark,” then that wouldn't work, because if “pig” was a bad word, then p!g and p ! g and p! g would also likely be in the filter. But then what would the forum filter tell you that the sentence is? Would it just cut off part of this bad word? But then this wouldn't make sense, because if the filter says that the bad part is contained in p! despite the fact that that string itself would not be caught by the filter, then that seems to be a contradiction. I don't think this solution is foolproof.

Sliverus wrote:

Za-Chary wrote:

Not to mention that the filter currently does not censor all bad words within all phrases in the first place, so the compromise wouldn't work as intended anyway. One can see this by the phrase “Occasionally I will say that sushi tastes great.”

I am not suggesting that any additional phrases should be added to the filter right now.

I'm not saying that is your suggestion. I am saying that filtering based on “subtly inappropriate” phrases will not 100% work. But that's a bit of a moot point since you have now moved on to a new idea.

Za-Chary wrote:

What counts as a sentence? If you mean “separated by a punctuation mark,” then that wouldn't work, because if “pig” was a bad word, then p!g and p ! g and p! g would also likely be in the filter. But then what would the forum filter tell you that the sentence is? Would it just cut off part of this bad word? But then this wouldn't make sense, because if the filter says that the bad part is contained in p! despite the fact that that string itself would not be caught by the filter, then that seems to be a contradiction. I don't think this solution is foolproof.

Interesting. But what about this: If the punctuation itself is part of the phrase that triggers the filter, then the sentence will “end” at the next punctuation mark that appears after the filtered punctuation mark. I'll give an example:

Imagine “rip” is a bad word. Now imagine a user attempts to say this phrase:

“You're just a little r!p and no one wants to be your friend!”

Because the first exclamation mark is part of the phrase that triggered the filter, it isn't considered the “end of the sentence”. The end of the sentence is then, by default, at the exclamation point after the word “friend”.

Sliverus wrote:

(#57)

Za-Chary wrote:

What counts as a sentence? If you mean “separated by a punctuation mark,” then that wouldn't work, because if “pig” was a bad word, then p!g and p ! g and p! g would also likely be in the filter. But then what would the forum filter tell you that the sentence is? Would it just cut off part of this bad word? But then this wouldn't make sense, because if the filter says that the bad part is contained in p! despite the fact that that string itself would not be caught by the filter, then that seems to be a contradiction. I don't think this solution is foolproof.

Interesting. But what about this: If the punctuation itself is part of the phrase that triggers the filter, then the sentence will “end” at the next punctuation mark that appears after the filtered punctuation mark. I'll give an example:

Imagine “rip” is a bad word. Now imagine a user attempts to say this phrase:

“You're just a little r!p and no one wants to be your friend!”

Because the first exclamation mark is part of the phrase that triggered the filter, it isn't considered the “end of the sentence”. The end of the sentence is then, by default, at the exclamation point after the word “friend”.

How about this: when you get the notification, just include a random amount of words (7-15 should work) to either side. Then you don't have to detect sentences, and it's not immediately obvious which word it is because it won't be in the center.

[small]non-member post, take whatever I say with a hint of salt and remember there's no authority behind it[/small]

kkidslogin wrote:

How about this: when you get the notification, just include a random amount of words (7-15 should work) to either side. Then you don't have to detect sentences, and it's not immediately obvious which word it is because it won't be in the center.

I think both of our solutions could probably work fairly well. I'll listen to what others have to say.

Sliverus wrote:

kkidslogin wrote:

How about this: when you get the notification, just include a random amount of words (7-15 should work) to either side. Then you don't have to detect sentences, and it's not immediately obvious which word it is because it won't be in the center.

I think both of our solutions could probably work fairly well. I'll listen to what others have to say.

yeah but what does this actually fix? isn't this just adding extra steps to finding the word you want to take out?

(⌐⎔_⎔)//This is Jack, he's been fighting these evil kumquats since early 2023!


(((( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́) ( •̀ᴗ•́))))