+- Hack Forums (https://hackforums.net)
+-- Forum: Hack Forums Official Information (https://hackforums.net/forumdisplay.php?fid=1)
+--- Forum: Site News (https://hackforums.net/forumdisplay.php?fid=2)
+---- Forum: Suggestions and Ideas (https://hackforums.net/forumdisplay.php?fid=134)
+---- Thread: Scoring toxicity levels of posts w/ machine learning - Useful feature for interaction (/showthread.php?tid=5653494)
* - Mαxim - Jun 23, 2017
(Jun 23, 2017 09:31 AM)Wαlrυѕ Wrote: Toxic levels are mainly measured by how they impact the conversation. A toxic level of 0%-10% is likely to be very beneficial and resourceful while a toxic level of 70%-100% means it is not contributing a valuable point to the conversation and possibly being negative to the conversation.
Your message got 34% likely to be perceived as "toxic" because it is not beneficial to the conversation. You are not displaying powerful points in your message. The message almost ridicules those who have a different opinion. This is something that you could debate, however generally people read comments like those as "arrogant sounding"
Fuck me this suggestion is stupid. How is my toxic level now?
* - Wαlrυѕ - Jun 23, 2017
(Jun 23, 2017 01:03 AM)SellingForums Wrote:
Not the best system.
![[Image: 514c65c3ab6dd4cdc9c76dba2302e65f.png]](https://i.gyazo.com/514c65c3ab6dd4cdc9c76dba2302e65f.png)
PerspectiveAPI measures how things impact a conversation, so this is not impacting the conversation negatively. You are displaying a point proficiently without offending people of another perspective. And keep in mind that some places have the age of consent at 16, so your point has been proven invalid.
(Jun 23, 2017 09:34 AM)Maxim Wrote: Fuck me this suggestion is stupid. How is my toxic level now?
Obviously a 100% because it is not affecting the conversation positively.
* - Mαxim - Jun 23, 2017
(Jun 23, 2017 09:36 AM)Wαlrυѕ Wrote: Obviously a 100% because it is not affecting the conversation positively.
Just wanted to test.
Welp, I think that it shouldn't be implanted now at least, doesn't seem to be advanced or smart enough at this stage.
* - Wαlrυѕ - Jun 23, 2017
(Jun 23, 2017 09:42 AM)Maxim Wrote: Just wanted to test.
Welp, I think that it shouldn't be implanted now at least, doesn't seem to be advanced or smart enough at this stage.
Yes it was only released to the public since 4 months ago. It's quite difficult to get training data, but I assume once they start partnering up with big social media/forum sites, than it will start becoming very proficient in understanding. Until then, I'll let this suggestion sit for a while and maybe in a few months we can come back to it
* - Charger - Jun 23, 2017
This is a forum, not a place to submit your essay.
Top 15 useless suggestions I've ever seen.
* - TyrantKingBen - Jun 23, 2017
Text analytics is cool but I don't see the usefulness for this to be on the front-end whatsoever. However, Omni could train a model to classify a shitpost and implement that in the back-end to help remove some spamming. I don't think it's worth the time or effort though.
* - X X X X - Jun 23, 2017
Interesting but I don't really like having it under every post. Maybe only visible for staff?
* - jujuger - Jun 23, 2017
(Jun 23, 2017 01:07 AM)Jendo Wrote: lmafooo
the system is not smart to understand the meaning of the sentences
it just has a list of toxic words, it checks if the sentence has any of them.
This. It doesn't under the blanket term of "machine learning", when it is obvious that the neural network it uses simply concluded "lol okay so if it contains words from database "negative", then it is bad".
It will not be able to detect sarcasm probably.
In fact, this is probably the dumbest idea I have ever heard. Why what if someone says something containing negative words that's IMPORTANT AND USEFUL, such as "warning: hackforums is currently under a ddos attack, please do not log in"
Or what if I am exposing a scammer on hf? "warning: Omni is a criminal and has scammed $9000 out of members with his ebook!"
Guys, sometimes "negativity" is useful depending on the social context.
In addition, why would i need a computer to tell me if something is negative? i can see clearly that "kill all jews" is negative and i do not need a computer to tell me this
* - Otis - Jun 23, 2017
(Jun 23, 2017 01:03 AM)SellingForums Wrote:
Not the best system.
I bursted out laughing when I saw this...I'm horrible lol
* - ra1d - Jun 23, 2017
What is this??? FamilyFriendlyForums.net?
Shit suggestion imo
* - +Cloud - Jun 23, 2017
(Jun 23, 2017 09:36 AM)Wαlrυѕ Wrote: PerspectiveAPI measures how things impact a conversation, so this is not impacting the conversation negatively. You are displaying a point proficiently without offending people of another perspective. And keep in mind that some places have the age of consent at 16.
Age of consent here is 16 so reading that was fucked for me it pretty much said i wanna have sex with someone under 16, your point is invalid
* - Wαlrυѕ - Jun 23, 2017
(Jun 23, 2017 05:53 PM)+Cloud Wrote: Age of consent here is 16 so reading that was fucked for me it pretty much said i wanna have sex with someone under 16, your point is invalid
It really is merely impossible for an algorithm to judge the opinion in a message and whether it is "messed up" according to society. It would take years before algorithms could inherit those abilities.
My main vision with Hackforums is for those who are being an asshole to realize that they are being toxic. Although it wouldn't do anything meaningful, it's a step in the direction of helping users post more HQ. Many users have been so LQ that Omniscient has banned them just based on one reply. Most, if not all, would have been rated a 100% using the PerspectiveAPI. I am not defending the system yet, as it clearly needs more work, however it should be something to consider in the future.
(Jun 23, 2017 05:01 PM)jujuger Wrote: Guys, sometimes "negativity" is useful depending on the social context.
In addition, why would i need a computer to tell me if something is negative? i can see clearly that "kill all jews" is negative and i do not need a computer to tell me this
One problem with negativity is that if used incorrectly, it negatively impacts the conversation as well. PerspectiveAPI has been very good at recognizing when someone is posting something without adding real content value.
(Jun 23, 2017 04:46 PM)X X X X Wrote: Interesting but I don't really like having it under every post. Maybe only visible for staff?
From what I have researched, the whole idea is based on theory. How it would be made in application would be up to if and how the idea was implemented.
Also notice many of the people who are against the idea are being toxic themselves.
* - thales - Jun 23, 2017
I mean... doesn't work well for tech talk.
![[Image: 3baf8bf614a472e8b9df2e4e5ff6abf8.png]](https://i.gyazo.com/3baf8bf614a472e8b9df2e4e5ff6abf8.png)
* - Wαlrυѕ - Jun 23, 2017
(Jun 23, 2017 06:12 PM)thales Wrote: I mean... doesn't work well for tech talk.
Sir... please scroll up and read what I have already said. It would be many years before an algorithm would be able to recognize these sorts of things, but as of now it's a work in progress.
* - X X X X - Jun 23, 2017
Took me a couple tries but I got this one :D
![[Image: fa6f3dc73fabc98195082a607eabe43d.png]](https://i.gyazo.com/fa6f3dc73fabc98195082a607eabe43d.png)