The last few days I have been learning a bit about machine learning. So I thought it might be a nice idea to develop a machine learning python script to locate comment abuse. It is still work in progress but I think I have it mostly thought out.
Photo taken in Santorini
I made a function which given some text checks for sentence similarity, change in vocabulary and a couple of other stuff, all of this information is summarised in a couple of numbers. These are something like a score. Having low scores does not mean that you are an abuser, similarly having high scores does not mean that you are not an abuser. So why calculate these scores? I expect that abusers will have similar scores. Abusers cluster together like flies cluster together on shit. :D I can identify these clusters using my scores and some math. In doing so, I have a good guess if someone is an abuser or not.
One of the drawbacks of this method is that abusers could actively try to get the right score. But for certain cases you can see that this is happening by tracking score changes of users.
Currently, I need to get all the standard steem pythons scripts functioning on Windows. I did have this functioning on my Linux machine but I had to return that machine to university since my contract expired. I am also still working on adding some more scores to my programs.
