For a post in Swedish directed at a Swedish general audience, read rather Noll Hat. This post is intended mainly for coders.
Dear people and in particular the parts of the audience stumbleing upon this site who happen to be from Sweden. None of you will have missed the awful display of hatred, bigotry and plain idiocy depicted in this weeks Uppdrag Granskning. This comes the same week that our region in the Nordic is haled by The Economist as role models. The grass, perhaps plenty green over here, is twiny, sharp and nasty.
In short, women are subjected to insults and death threats and everything in between, when they express their views on Facebook.
Amid outcries and shame, the pragmatist within me caught foothold on something. H&M say they deleted 3000 comments. 2000 remained but that’s beside the point. It takes too long to read through, monitor and delete posts on your Facebook pages. I’m guessing max 10 per minute per person. Comments stream at many times that speed, at least when it counts the most i.e. most people are reading. It’s clear to me that on trendy topics the rate of commenting, fair or otherwise, is a critical phenomenon and the growth is exponential. Certain subject attract women haters and they spur each other to write ever more twisted comments.
We need better tools to curb the rate, early and effectively. I tweeted about it and at the same time realized it was within my capacaity to do it and fairly quickly.
The result is Zero Hate.
This rather ruddy proof of concept lets you select multiple tweets and deletes them in batch. It’s really fast and sure beats doing it one by one.
It’s of limited use unless you can sort the comments according to contents of certain words and that’s the primary target of the upcoming version.
The following this is the first screencast I’ve made and I know it shows. But it got the word out. It’s in Swedish.
So this is a call for participation. Tweet me.
Anyone who can take over the server coding would be most welcome. Any platform will do.
Anyone who has experience in natural language processing, let thyself be known. We need to process comments server side and score them while training the AI, only temporarily persisting them.
Feature request list:
- Selectable pages
- Search for keywords, rank and order comments
- Add to and persist, keyword list
- Locally persisted comment store
- Support for Facebooks paging, field limitation
- Batch deletes
- Realtime updates
- Language processing/machine learning