• Please take a moment to look over the Suggestions & Feedback rules before making a post: READ RULES HERE

AI for BHW

healzer

healzer

Elite Member
Joined
Jun 26, 2011
Messages
2,974
Reaction score
3,063
Hi all, :)

Once in a while I open BHW to see what's happening.
This morning I saw BHW was full with spam threads, at least 20-40 duplicate threads trying to promote some dating site.

I am thinking of building an AI tool that recognizes spam threads and even prevents them from being created. It also allows us to automatically flag/report and prevent posts that violate BHW rules.

Such a system can be built by leveraging the community to train it (human assisted learning).
The ideal outcome would be to reduce workload for moderators/admin, such that they can instead focus on more meaningful work.

What do you think?
 
Advertise on BHW
Am sure it would be really helpful one for site mod, as long My self I have seen many threads promoting their service with mobile numbers and website URLs which it is against the bhw rules . The tool that you are planing to build that should control even from posting the content on site .
If it is worked then it’s great one I say . Any way congrats @healzer
Have a nice day
 
Sounds great. Hope it works and that the Admins will implement it.
 
2A12DB8E-C720-40F4-9E7A-CA3645F2CDBC.png

Apparently from my knowledge it’s been happening for years. It happens on Reddit too!
 
People have given them solutions in the form of regex and even screenshot the forum setting to implement this and I think the site owner said it's not needed.

So going with AI over regex seems a bit pointless and a waste of time, perhaps other forums that actually prevent spam instead of encourage it would buy your idea as a pluggin. Marketing it to multiple websites for free/paid would also give you much more data to make accurate decisions with your machine learning rather than just implementing it on this forum.
 
healzer said:
Hi all, :)

Once in a while I open BHW to see what's happening.
This morning I saw BHW was full with spam threads, at least 20-40 duplicate threads trying to promote some dating site.

I am thinking of building an AI tool that recognizes spam threads and even prevents them from being created. It also allows us to automatically flag/report and prevent posts that violate BHW rules.

Such a system can be built by leveraging the community to train it (human assisted learning).
The ideal outcome would be to reduce workload for moderators/admin, such that they can instead focus on more meaningful work.

What do you think?
Click to expand...
Should be easy enuff if somebody wanted to try. For a head start, may be look into "Naive bayes classifier". ;)

carmamag said:
People have given them solutions in the form of regex and even screenshot the forum setting to implement this and I think the site owner said it's not needed.

So going with AI over regex seems a bit pointless and a waste of time, perhaps other forums that actually prevent spam instead of encourage it would buy your idea as a pluggin. Marketing it to multiple websites for free/paid would also give you much more data to make accurate decisions with your machine learning rather than just implementing it on this forum.
Click to expand...

Regex does not learn, AI does. Calling AI "pointless" doesn't make sense lol. If you manage to block the spammer with the regex today, hewill come up with something passable tomorrow. If the regex solution was correct, then giants like google, fb or insta won't be implementing AI anyway.
 
Last edited:
Gogol said:
Regex does not learn, AI does. Calling AI "pointless" doesn't make sense lol. If you manage to block the spammer with the regex today, whe will come up with something passable tomorrow. If your solution was correct, then giants like google, fb or insta won't be implementing AI anyway
Click to expand...

All the spam threads contain phone numbers in the title/content

Regex the phone number as @MatthewGraham's thread showed and the spam will stop

@Diamond Damien said this would give too many false positives, except when would anyone NEED to post a phone number on here in a title? Regex would only match the phone number so I don't know what the false positive would be.

I have degree in comp sci, I know the difference between regex and machine learning.
 
We are not only talking about phone numbers ...
 
healzer said:
Hi all, :)

Once in a while I open BHW to see what's happening.
This morning I saw BHW was full with spam threads, at least 20-40 duplicate threads trying to promote some dating site.

I am thinking of building an AI tool that recognizes spam threads and even prevents them from being created. It also allows us to automatically flag/report and prevent posts that violate BHW rules.

Such a system can be built by leveraging the community to train it (human assisted learning).
The ideal outcome would be to reduce workload for moderators/admin, such that they can instead focus on more meaningful work.

What do you think?
Click to expand...

Just like CleanTalk?
 
carmamag said:
I have degree in comp sci, I know the difference between regex and machine learning.
Click to expand...
Well, then don't compare regex with AI ;)
carmamag said:
All the spam threads contain phone numbers in the title/content
Click to expand...
carmamag said:
Regex the phone number as @MatthewGraham's thread showed and the spam will stop
Click to expand...
"Do not trust your user input", is what I was taught by my first "web development teacher". I still follow that to this day. AI works pretty well for spam filtering (you can read up more on this here). @Diamond Damien was totally accurate when he said "it will give false positive". If you block a certain number, the spammer will just come up with a new number. If you block all numbers, you are gonna create a lot of trouble.

AI would be a lot better option because the system "learns" from the dataset provided instead of blindly trying to block everything. I am not saying AI is bullet proof, I am just saying that regex is a laughable solution.
 
Gogol said:
Regex does not learn, AI does. Calling AI "pointless" doesn't make sense lol. If you manage to block the spammer with the regex today, hewill come up with something passable tomorrow. If the regex solution was correct, then giants like google, fb or insta won't be implementing AI anyway.
Click to expand...
If you search for the terms that are being spammed, BHW isn't the only site being hit. Spam targeted in bulk at arbitrary sites is rarely adjusted for individual cases.

carmamag said:
All the spam threads contain phone numbers in the title/content

Regex the phone number as @MatthewGraham's thread showed and the spam will stop

@Diamond Damien said this would give too many false positives, except when would anyone NEED to post a phone number on here in a title? Regex would only match the phone number so I don't know what the false positive would be.

I have degree in comp sci, I know the difference between regex and machine learning.
Click to expand...

I did test the regex against some test threads to see if there would be false positives. If you match 9-20 characters, you're basically only matching phone numbers.
  • /[0-9-()#. ]{9,20}/gi
Can also do this and only match 91 numbers, which is what most of the spam is for (India country code).
  • /91[0-9-()#. ]{9,20}/gi
And then also set it up to only apply to users with under a specific post count (e.g., users with under five posts). Should be almost 100% accurate with virtually no collateral damage to real posts.

Anyway, not to micromanage the forum, since I don't work here. Since this topic came up again, there's some more detail.

Original thread: https://www.blackhatworld.com/seo/0-9-0-9-3-s-0-9-3-s-0-9-4-6-ig.1089276/
 
Also, this kind of text analysis is actually a good fit for machine learning.

Although, an AI spam filter for a single forum would be at best a questionable use of resources. It makes way more sense to create an AI spam filter as a plugin or similar and to distribute it for use publicly. The process would be fairly time consuming and you'd be able to benefit from a much larger data set that way.
 
reminds me of this meme
deep_learning.jpg
 
Let's make everyone a MOD to delete spam threads. I know I'm old school, so don't remind me of.. lols!
 
MatthewGraham said:
Also, this kind of text analysis is actually a good fit for machine learning.

Although, an AI spam filter for a single forum would be at best a questionable use of resources. It makes way more sense to create an AI spam filter as a plugin or similar and to distribute it for use publicly. The process would be fairly time consuming and you'd be able to benefit from a much larger data set that way.
Click to expand...
This is exactly the plan, however, we gotto start somewhere right.
 
there was a challenge on Kaggle. Detecting Hate Speech on Twitter I believe. This is something similar I guess
 
Advertise on BHW
You must log in or register to reply here.
Sign up now!

Main Menu

Home Main Forum List Black Hat SEO White Hat SEO BHW Newbie Guide Blogging Black Hat Tools Social Media Downloads

Marketplace

Account Selling Content / Copywriting EBooks / Methods Hosting Misc Proxies For Sale SEO - Link building SEO - Packages Social Media Social Media - Panels Web Design More

Making Money

Affiliate Programs Hire a Freelancer IM Journeys Making Money Pay Per Click Site Flipping Want to Buy

BlackHat World

Dispute Resolution Forum Suggestions & Feedback Introductions Lounge
Back
Top