With detoxifAI, Mitigate Cyberbullying

A transparent, de-biased toxic comment detector that gives live feedback

Why detoxifAI?

detoxifAI is the only AI-based Toxic comment detection tool that is not only vetted for bias, but is also is transparent on toxic terms - helping children and teenagers understand better what makes a comment toxic.

For Parents

Learn more about how detoxifAI can protect your child against online abuse.

Schools

For School Administrators

Don't leave your school legally responsible for bullying. Learn how detoxifAI can help.

Forums

Online Forum Administrators

Customize your online forum with detoxifAI and ensure that your users follow community guidelines without using toxic language.

The Tech

detoxifAI uses state-of-the-art Artificial Intelligence to ensure that you get accurate, transparent results.

Hierarchical Attention Network

detoxifAI uses an advanced Hierarchical Attention Deep Learning Neural Network, allowing it to not only predict if a comment is toxic or not but also enabling it to be transparent and indicate why a comment could be toxic.

Bias Mitigation

detoxifAI uses the neural network output, linguistic insights and a grid search algorithm to detect identity terms that are used in a toxic way and to remove the potential biases.

Accuracy

detoxifAI achieves a test accuracy of 98% with a bias reduction of over 44%.

Automation

detoxifAI is built to quickly detect toxic comments while also suggesting possible source of toxicity that can help a user correct their behavior


Experiments

Below are some samples from our experiments. The highlighting indicates the intensity of the toxicity.

You shut your mouth. And my contributions will never be missed. As long as there are several hooligans on this site, I will remain to sort these issues out the proper way. Currently, all attempts have failed. But that won't be the case for much.
Model Prediction: Toxic
you shut your mouth. and my contributions will never be missed. as long as there are several hooligans on this site i will remain to sort these issues out the proper way. currently all attempts have failed. but that won't be the case for much longer. permanent blocking won't work with me as my comeback will be uplifting.
An overwhelming share of America’s lesbian, gay, bisexual and transgender adults (92%) say society has become more accepting of them in the past decade and an equal number expect it to grow even more accepting in the decade ahead.
Model Prediction: Non-Toxic
an overwhelming share of america's lesbian gay bisexual and transgender adults say society has become more accepting of them in the past decade and an equal number expect it to grow even more accepting in the decade ahead .
Why is this so hard to understand? Are you out of your mind, you idiot!
Model Prediction: Toxic
why is this so hard to understand? are you out of your mind you idiot!
You are really stupid and dont deserve to be here.
Model Prediction: Toxic
you are really stupid and dont deserve to be here .

Live Demo



The Live Demo uses machine learning models to classify comments as toxic or nontoxic. Disagree with the result? Please let us know using the Contact Us page. While our model is robust, we are still are sill in our infancy and we may get some things wrong.

Get detoxifAI Today.

Install the detoxifAI Chrome Extension Today and protect your loved ones from online abuse.

Get Chrome Extension

Contact Us

Your message has been sent