Let's build an insultbot

Online Content Moderation with AI: A Practical Guide

The challenge of online content moderation is significant. Determining what content should be visible and what needs to be moderated or removed is a complex issue. Even prominent figures have experienced it firsthand – the Prime Minister of Norway herself had a photo she posted on Facebook quickly removed.

This course explores a practical application of artificial intelligence for online content moderation, leveraging the power of TensorFlow.js – machine learning in JavaScript. We'll utilize a pretrained model to identify content that should not be displayed in conversations. The beauty of this approach is its ease of use: simply call the model and receive a list of content types flagged for blocking.

The need for effective and safe online content moderation is constantly growing. While sophisticated approaches are being explored, including advanced sentiment analysis with tools like ChatGPT, this course offers a free alternative that requires less coding expertise. ChatGPT, while powerful, can incur costs with its API calls.

We'll focus on a common yet challenging case: direct insults (name-calling). The solution employs a classical pattern recognition approach, assigning a probability score to identify potentially offensive language. This probability, a standard in deep learning, allows you to define your own tolerance level for insults.

This course utilizes two powerful, MIT license Google public tools: TensorFlow.js and Angular. TensorFlow.js is a versatile deep learning tool for building intelligent web applications. Angular provides a free framework for creating robust frontend applications, eliminating the need for server-side development. Learn how to implement a content moderation system using these accessible technologies.