Hello Scratch Team!

I have a suggestion! I've tried to submit a forum post in the “Show and Tell” section of the forum, but the algorithm falsely detected unsuitable language. I've re-read my message and scanned for accidental combinations (since the bot reads the content without spaces) and found none. The message that occurs is:

Sorry, this post appears to include unsuitable language and will not be posted.

Especially for longer messages, I'm suggesting a feature where it highlights the parts of your message where it detects “unsuitable language”, so that people who didn't intend to type something bad know what to fix. This will save a lot of time and frustration among users who just wanted to post something without getting attacked by the bot!

Thanks!
- LeggoMyEggos123

this is definitely a dupe ima go find the main topic brb

Strong agreement. This should help people satisfy the bot while being able to continue posting as normal.

ok I can’t find one because the dupe finder thing doesn’t know how to find something relevant to what I’m searching. I guess this can stay up until someone finds the older topic

There is one big problem with this - if the indicator points out a word that may be inappropriate, now the person knows about the inappropriate word. It's basically just teaching people inappropriate words, even though the person never may have intended to write the word in the first place. Let's say I wrote this:

fridays are cool

Okay, let's say “day” is a bad word. Note that I never actually intended to write “day” as a bad word and I didn't even know about that word in the first place. The indicator would probably look like this:

fridays are cool

Well now I know that “day” is a bad word because of Scratch. Which isn't okay.

SidewaysCoder wrote:

There is one big problem with this - if the indicator points out a word that may be inappropriate, now the person knows about the inappropriate word. It's basically just teaching people inappropriate words, even though the person never may have intended to write the word in the first place. Let's say I wrote this:

fridays are cool

Okay, let's say “day” is a bad word. Note that I never actually intended to write “day” as a bad word and I didn't even know about that word in the first place. The indicator would probably look like this:

fridays are cool

Well now I know that “day” is a bad word because of Scratch. Which isn't okay.

Not the best example…a better one:
Let's say “cat” is strong language.
Now let's see this:

The catcher hit the ball

I try to send it, but it highlights

The catcher hit the ball

Well, now we know “cat” is strong language, which could lead to Scratch getting sued by overprotective parents.

So anyway…
How about if we highlight the entire section without spaces, instead of the problem itself?

The catcher hit the ball

Now we can fix the problem without the kid knowing the strong language.

@SidewaysCoder That's a major flaw I haven't even thought of, thanks for pointing that out! There could be a similar alternative solution, too. For example, if “sus” was a bad word and you typed:

"Among Us is sus"

The bot could say something like:

An inappropriate word was detected at line 1

That way, people who know what “sus” means can correct the mistake. Even if someone doesn't know what it means, the context of “line 1” could help them rewrite the single line of their message, instead of having to randomly guess what the bot flagged as inappropriate (and fixing multiple words and phrases that weren't the culprit, without the person learning what “sus” means).

Or, there could be a change in algorithm to prevent these mistakes as well. Like, if the words “was” and “still” were typed with a space between them (of course without spaces would contain a certain offensive word), the bot could learn to recognize that word wasn't intentional and make an exception.

Edit 1: @TheCreatorOfUnTV, just saw your idea to fix that problem after I posted this response. That's another good idea!

TheCreatorOfUnTV wrote:

SidewaysCoder wrote:

There is one big problem with this - if the indicator points out a word that may be inappropriate, now the person knows about the inappropriate word. It's basically just teaching people inappropriate words, even though the person never may have intended to write the word in the first place. Let's say I wrote this:

fridays are cool

Okay, let's say “day” is a bad word. Note that I never actually intended to write “day” as a bad word and I didn't even know about that word in the first place. The indicator would probably look like this:

fridays are cool

Well now I know that “day” is a bad word because of Scratch. Which isn't okay.

Not the best example…a better one:
Let's say “cat” is strong language.
Now let's see this:

The catcher hit the ball

I try to send it, but it highlights

The catcher hit the ball

Well, now we know “cat” is strong language, which could lead to Scratch getting sued by overprotective parents.

So anyway…
How about if we highlight the entire section without spaces, instead of the problem itself?

The catcher hit the ball

Now we can fix the problem without the kid knowing the strong language.

Perfect

Both are better solutions, but @TheCreatorOfUnTV's solution still makes it easier for the kid to learn bad words since it's still too specific, and both solutions will still be confusing to someone who already looked through and can't find the bad word because they don't know the bad word triggering it, especially in @LeggoMyEggos123's solution if there are lots of words in that single line (and it would be way too specific if the line only had 1 word).

Definitely support this, I’ve had this happen to me multiple times, with me having to randomly delete things because idk what the bot wants. Turns out the bot can even think my own username is bad! The bot is wild sometimes lmao

While I wish this could happen, our two options are either accidentally have kids discover bad words through the bad word detector or make it more vague which doesn't exactly help much.

That said, they seriously need to rework the bad word detector. The fact that my signature is real says something. I think maybe a better idea would be an option to report false positives if you can't find anything wrong so that the Scratch Team can improve things.

Type hello and it thinks it is the H word, wait - it detected the o

This is similar to this rejected suggestion:

tolors wrote:

6.5 Detailed information on why a comment was filtered
Some Scratchers have suggested that the filter should be more clear by showing which letters/words tripped the filter. However, such a suggestion would make it easier to write comments that are disrespectful or inappropriate but do not trip the filter. When a Scratcher is prevented from posting a comment due to the filter, the Scratch Team wants to encourage them to think carefully about why, so that they can refrain from writing similar comments in the future.

WigglyJoey wrote:

This is similar to this rejected suggestion:

tolors wrote:

6.5 Detailed information on why a comment was filtered
Some Scratchers have suggested that the filter should be more clear by showing which letters/words tripped the filter. However, such a suggestion would make it easier to write comments that are disrespectful or inappropriate but do not trip the filter. When a Scratcher is prevented from posting a comment due to the filter, the Scratch Team wants to encourage them to think carefully about why, so that they can refrain from writing similar comments in the future.

Thanks for the link! It does seem like this is a rejected suggestion. I will close this topic to highlight the answer and prevent future responses.