My suggestion today is pretty straightforward- in a comment, before the filter runs, all special letters should be replaced with normal letters.

- Why?
Special text like tiny caps can evade the filter.

This is a huge problem, and it has been exploited before. The filter only censors text which is normal, so it doesn't detect special letters. This means I can say things like “bit challenging” using special letters and it will go through.

- How?
I propose that there 26 lists- one for each letter of the alphabet. In each list is the special text for a letter, like shown below.

If a special letter is found in the list for it's respective letter, it is replaced with the normal letter. This prevents people from abusing the filter.

- Pros and cons
Pros:

• This would prevent the filter from being abused, and would make scratch a lot easier to moderate.
  

• The moderators wouldn't have to blacklist every variation of a word, they could just blacklist the normal letters. Less work, better results.
  

• Spammers would need to get something else to spam about, as they typically use the special text exploitation.
  

Cons:

• People wouldn't be able to use special text, which would cause a huge backlash from the community. However, security is more important.
  

• Would require a lot of initial manual work. I think it's worth it when you consider the benefits of this suggestion.
  

edit:

blablablahello wrote:

what if the filter just does the replacement while filtering, but after filtering, still sends ou the message WITH the special font? that wouldn't really be prpoblematic would it?

I am aware of this solution, but I decided not to include it as it already takes lots of time, effort, and money for the ST to implement a suggestion.

Anyone is welcome to make a suggestion where this solution is implemented alongside what I've proposed.

I think it should be added, too many people abuse the filter like this

(long unnecessary quote removed by moderator - please don't spam)

OK, won't do that again, just figured out that rule now

.

Good idea. Knowing the community, they would definitely get mad at the ST for this. But it would definitely help, that's for sure.

support, google had something similar where if I search apple with a fancy font, it searches for apples like as if I didn't search with one

most scratchers here use fancy font just for fun, so semi support?

The lack of ability to use special characters (which should really only be limited to letters) is fine, they mess up screen readers anyway. You don't need to be fully blind to need a screen reader; I've known plenty of people who use it since they're way better at processing spoken word over written word, alongside other similar reasons.

The only issue is that these characters are in Unicode for a reason. I don't know all of them, but there are quite a few that exist for use within mathematics–to differentiate x's in “solve for x” versus “x marks the spot.” A simple filter would completely stop that, and there might need to be some checks to see if they're being used as intended or for other purposes.

That's the only edge case I can think of, though. Mostly full support here.

what if the filter just does the replacement while filtering, but after filtering, still sends ou the message WITH the special font? that wouldn't really be prpoblematic would it?

(misread op - I thought it was something like what the above post said)

blablablahello wrote:

what if the filter just does the replacement while filtering, but after filtering, still sends ou the message WITH the special font? that wouldn't really be prpoblematic would it?

Yes this would remove a con

33emeries wrote:

The only issue is that these characters are in Unicode for a reason. I don't know all of them, but there are quite a few that exist for use within mathematics–to differentiate x's in “solve for x” versus “x marks the spot.” A simple filter would completely stop that, and there might need to be some checks to see if they're being used as intended or for other purposes.

I envision the lists as something that is manually updated, so if it's something that is used in math the ST can manually remove it from the replacement list. This wouldn't apply to numbers anyway, only letters, and english ones at that too.

How would the filter for the fancy letters be able to detect inappropriate words before turning everything into regular text? (Like yesterday’s Griffpatch spam disaster.)

Warriorcatgirl2009 wrote:

How would the filter for the fancy letters be able to detect inappropriate words before turning everything into regular text? (Like yesterday’s Griffpatch spam disaster.)

it doesnt- it converts the special text into normal words, which would then set off the filter. The normalisation process runs before the filter runs.

Warriorcatgirl2009 wrote:

How would the filter for the fancy letters be able to detect inappropriate words before turning everything into regular text? (Like yesterday’s Griffpatch spam disaster.)

It detects them after turning it into regular text.

I also made a saying auto mod mute copy and paste or go to website and copy and paste project:
https://scratch-mit-edu.ezproxyberklee.flo.org/projects/812200221/
It also has a testing site!!,
https://scratch-mit-edu.ezproxyberklee.flo.org/studios/32954944/comments

Sorry if I am advertising but I know it is related

blablablahello wrote:

what if the filter just does the replacement while filtering, but after filtering, still sends ou the message WITH the special font? that wouldn't really be prpoblematic would it?

Good idea! I was about to post this, actually.

bump

up it goes, don't let this suggestion fall into the same fate as https://scratch-mit-edu.ezproxyberklee.flo.org/discuss/topic/655895/

Support. Scratch is a kids website, working around the filter with special text should be fixed so there are no bad words. Also, what if the filter only converted the special text into normal for the filtering step, but then back into the special text?

Support, because filtering. However, if it detects Latin characters, it should automatically ignore any LTR and RTL characters when scanning.

Support, this would improve the filtering system in obvious ways.