Update of the Steem Sincerity feature

in #steemdev6 years ago (edited)

After releasing the Steem Sincerity Anti-Spam integration to SteemPlus, I have collected your feedback from the comments and decided to make a small adjustment. These will

Although I made pretty clear that the main purpose was to help training the machine learning algorithm, some have voiced concerns about the false positives in the Bot and Spammer categories. Our intention was never to discriminate these accounts but rather to let people give feedback to correct those errors.

This being said, we understand that being classified as a spammer when you are not can be frustrating so we decided to add a new category "Pending" used when the algorithm is unsure of the result.

This is how it will be done :

  • If the biggest percentage is attributed to Human category, the user is considered Human
  • If the biggest percentage is attributed to Bot category and is over 80%, the user is considered as a Bot
  • If the biggest percentage is attributed to Spammer category and is over 80%, the user is considered as a Spammer
  • Otherwise, the user category is set as Tell us, and other users are welcome to rate this account.

A more than 80% attribution to those categories will find strong similarities with manually selected Spammers or Bots and will prevent having the many false positive that we currently have.

It also mean that we will have a lot of Tell us in the first place.

Therefore, we count on you to help us classify all those Pending accounts!

@stoodkev for @steem-plus

Sort:  

Glad I'm still a "human" {beep}

An excellent update. Really it is very useful. I am going to vote you as a witness. Around 60 community members selected me as their proxy for witness voting. I hope it will be much helpful to encourage like this good updates and services. At present, I am working on my tab. Once i reach my home and login in my pc, i will vote your as witness. Thanks once again for this much of good service.

Thanks a lot! Looks like you forgot to cast the vote though ^^

wow great article..
Thanks for this ..
I learnt from it especially when you talked about
"how the main purpose is to let people give feedback to correct those errors".

If an given the opportunity, I would love to be path of it. To impel see it to a better place.
Looking forward to seeing more of your posts.

Thanks for your work. I appreciate it.

really, i must say this is amazing. but really, when will this be made available for mobile devices?

I saw that new feature pop up yesterday and was a bit distressed to find I'm 28.57% spammer! How did I get this and how do I exorcise it?

Machine learning use a statistical approach, it doesn't matter that you re 28.57% spammer as long as you re more human. Some behaviors are similar in both categories, hence the result.

my brother, @lefthouse was marked as a spammer but i know for a fact he is not. he is my brother. a human! can you help sort him to human status? thanks kindly!

Good idea and has been done very perfectly, let this katagore is very important here. I have complained for a long time about the number of spamms that haunt me on every post and comment. If allowed I can be part of this project.

I don't understand the classification.
clever, cleverbot, banjo and catfacts are classified as "human"?

In other words, that if human, most of the content will be classified as such, and if bot, then actions are more than not often limited.

But clever, cleverbot, banjo and catfacts are definitely not human.

Then the algorithm needs some help and you need to vote on those bots that are automatically posting. All it takes are two simple clicks and it benefits the community for you to help the algorithm learn more about commenting/posting habits of humans and bots.

Thanks a lot for getting it @idontgnu !

I didn't think it was that difficult, pretty straightforward!

Coin Marketplace

STEEM 0.30
TRX 0.12
JST 0.036
BTC 66992.58
ETH 3460.53
USDT 1.00
SBD 4.68