/gaali

Toxicity Detection

Primary LanguagePHP

Gaali

Gaali removes harmful and unwanted comments before they can cause any damage. Its checks for toxicity based on TensorflowJS.

Description

Gaali in Hindi means Slang which is Tocicity in today's jargon

Stop toxicity and form slang - Gaali effectively eliminates harmful and unwanted comments, preemptively preventing any potential damage.

Frequently Asked Questions

  1. After posting a comment, why does it still say it's in moderation ? Isn't this supposed to be automated ?

Yes, it's automated and there is no need for the wordpress admin to approve a good comment. But to validate a comment for foul language and slangs, which is called a toxicity check, it takes time, like many many seconds, to detect it.

  1. Why does it take so much time for a toxicity check ?

The toxicity check connects to TensorFlow, here using TensorFlowJS, which is a Google's product and the data sets used are on Google's servers.

  1. So this connects to some 3rd party backend-service ?

Yes, it connects to an API at toxicity.co.in which is again, my (free) service which runs on Deno at Deno Deploy - deno.com which in turn goes to Google for returning the results of toxicity. Sometimes it takes a minute or more to return the result and at times, it even times out. If it times out the admin has to manually validate the comment in wp-admin.

  1. Will this plugin remain free forever ?

I want it to remain free forever and open-sourced. There is so much of free stuff on the internet and even open-sourced. I want to be part of the community and don't intend on commercializing it. But if toxicity service at my deno.com account take a hit on many requests, I will then have to shell out minimum $10/mo ($2 per million requests per month, $0.30/GiB outbound data transfer) Even then that's okay and I can fund that minimum amount, but if it goes beyong $10 a month, then I will seek funding of some sort.

Screenshots

  1. This is how the comment list in the backend looks with details of the toxicity parameters :

Comments Admin

Changelog

= 0.5 =

  • Just started, will update when I come across issues.

Upgrade Notice

= 1.0 = This is the first version.

Features

  1. Comment check runs in the background
  2. Neat table in comment's admin showcasing 6 toxic parameters and a master parameter Toxicity that will decide if the comment should be auto-approved or not
  3. API Key not required for now, but will incorporate this if there are many users of this plugin and the usage is high