News for hackers. Stuff that matters.
Re:Are popular toxicity models simply profanity detectors? by mrr54 on Monday January 24, 2022 @10:38PM (#30067646)
If I could filter out the overuse of profanity shown in this article, I would. "Fuck yeah!!! That bad bitch is totally the shit!!" gets caught by a profanity filter. No great loss, IMO.

If you get normal not-always-online not-gen-Z people to evaluate these messages and label them as Good or Bad then you will get results like this. If I got any member of my family over the age of 30 to evaluate these messages, they'd label them all as offensive.

Re:Are popular toxicity models simply profanity detectors? by Talanes on Monday January 24, 2022 @11:31PM (#30068035)
>If I got any member of my family over the age of 30 to evaluate these messages, they'd label them all as offensive.

How sure are you on that? I'm 31 and there's definitely no one in my peer group who would find those examples offensive. Hell, only 1/2 parents might be offended by that language, and my mom has been working around young people enough for the past decade that she might have softened up since I lived with her too.

Re:Are popular toxicity models simply profanity detectors? by jayd16 on Monday January 24, 2022 @10:49PM (#30067724)
Pretty sure anyone who lived through the 90s should be able to spot the difference between shit and the shit. MJ released Bad in 87.
Re:Are popular toxicity models simply profanity detectors? by mrr54 on Monday January 24, 2022 @11:27PM (#30068003)
Calling something "the shit" is still unnecessarily using profanity. It's impolite. It's not a surprise it would be flagged as offensive.
Re:Are popular toxicity models simply profanity detectors? by tablespoon on Tuesday January 25, 2022 @12:10AM (#30068332)
> Calling something "the shit" is still unnecessarily using profanity. It's impolite. It's not a surprise it would be flagged as offensive.

It's not impolite, it's just very informal.

Re:Are popular toxicity models simply profanity detectors? by riversflow on Tuesday January 25, 2022 @12:03AM (#30068279)
Strong disagree from me. I don't think it's impolite at all. I find your attitude in favor of what I would consider cultural erasure offensive.
Re:Are popular toxicity models simply profanity detectors? by newbytuby on Monday January 24, 2022 @11:43PM (#30068113)
But these are all subjective opinions – I could just as easily say that it _is_ a surprise to me that it's flagged as offensive, or that unnecessarily using profanity isn't toxic or impolite. And I feel like that's where this post is gesturing toward – not a judgment on what should or should not be considered toxic, but just a reminder to be intentional when writing definitions and sourcing training data.
Re:Are popular toxicity models simply profanity detectors? by xboxnolifes on Tuesday January 25, 2022 @01:14AM (#30068751)
Maybe it is to a quickly shrinking part of the population, but it's basically common speak to anyone in my life under the age of 45.
Re:Are popular toxicity models simply profanity detectors? by smaudet on Tuesday January 25, 2022 @01:38AM (#30068870)
Chiming in to say you are dearly mistaken, you turkey-squid uncle buntler.

I, on the other hand, am a fucking amazing shit OG bitch.

But yeah this stuff is absolutely dangerous, opinions like these are destroying culture. It won't be long until the threads these automatic tools are moderating, WONT BE WORTH HAVING.

'Nuff said.

Re:Are popular toxicity models simply profanity detectors? by pessimizer on Tuesday January 25, 2022 @12:21AM (#30068417)
It's not necessary to post at all. Necessity isn't the criterion.
Re:Are popular toxicity models simply profanity detectors? by tracker1 on Tuesday January 25, 2022 @12:07AM (#30068318)
Context matters.
Re:Are popular toxicity models simply profanity detectors? by vba616 on Tuesday January 25, 2022 @01:13AM (#30068746)
I first heard "the shit" meaning "good, cool, amazing" in 1996, I think. Never knew the origin, I heard it from a Californian in Arizona, and not really from anyone else since.
Re:Are popular toxicity models simply profanity detectors? by kimixa on Monday January 24, 2022 @11:38PM (#30068073)
The issue is more "Offensive" depends very much on context - not a black and white state.

There are contexts that quote would not be offensive to anyone (more casual settings with a group that trusts each other), but also contexts where everyone would find it offensive, no matter their age or generation.

Re:Are popular toxicity models simply profanity detectors? by perryizgr8 on Tuesday January 25, 2022 @05:49AM (#30070415)
> The issue is more "Offensive" depends very much on context - not a black and white state.

Agreed. As an example, your usage of "black" and "white" in this manner could be offensive to someone in some context.

Re:Are popular toxicity models simply profanity detectors? by politician on Tuesday January 25, 2022 @01:35AM (#30068853)
You need to train models on everyone's own labels. As you've said, "offensive" is context-dependent and each listener is huge part of that context.
Re:Are popular toxicity models simply profanity detectors? by StanislavPetrov on Tuesday January 25, 2022 @01:55AM (#30068957)
Which is why its futile and counterproductive to have content-based mass censorship. Offensiveness is completely subjective. Give everyone a mute button and a blacklist for them to block out any posts with whatever whatever words they are too fragile to see and let people decide for themselves. On large social media platforms it would be far better if we went back to the old system of banning spammers and people posting explicitly illegal content and letting everything else sort itself out.
Re:Are popular toxicity models simply profanity detectors? by pjmlp on Tuesday January 25, 2022 @01:54AM (#30068945)
Reaching 50 over here, I don't see them as offensive, in fact they are quite light.
Re:Are popular toxicity models simply profanity detectors? by maxerickson on Monday January 24, 2022 @11:38PM (#30068079)
The point the article is making is not that the messages are "Good" or "Bad", it's that an automatic rating system that intends to allow some usages of profanity rates messages with the type of profanity that it purports to allow as very bad, towards the end of the scale that it uses.
Re:Are popular toxicity models simply profanity detectors? by AussieWog93 on Tuesday January 25, 2022 @02:07AM (#30069027)
Where are you from, and what's your cultural background? I'm from Melbourne, and my parents are from Adelaide (possibly the most puritan city in Australia).

Not a single person I know would take offense to this.

My South African wife's family, however, is a completely different story. Her mother's side (English) would be horrified by this, whereas her dad's side (Afrikaner) wouldn't be upset in the least.

Point is, it's geography and upbringing far more than age that predicts puritanism. If anything, the young people from your area are regressing to the global mean.

Re:Are popular toxicity models simply profanity detectors? by hoseja on Tuesday January 25, 2022 @03:05AM (#30069401)
Sounds like a Mormon or some other form of American.
Re:Are popular toxicity models simply profanity detectors? by raverbashing on Tuesday January 25, 2022 @02:09AM (#30069048)
"overuse of profanity"

Not everything is the Disney Channel and not everybody shares your views on "morality"

I actually think that too saccharine language is trying to hide something, so immoral.