FAIRMod: Detecting & Restricting Harmful Content using Machine Learning

The Content Restriction System (CRS) harnesses the power of machine learning to mitigate the proliferation of harmful content in digital spaces. The current systems are mostly made by private social media giants and there is a lack of a wholesome all-round CRS for all. This can be useful to remove bad comments in the comment section, removing it from the reach of children as well. It detects profanity, harmful and violent behavior. We are creating this using Long Short Term Memory (LSTM) and the algorithm is modified such that it detects the comment as quickly as possible. This is an improved version of this algorithm that can give us faster results. For this we have implemented several filters before sending a text object for analysis minimizing the need for analyzing all the data objects. After these filters are applied, we send the data objects for analysis and even this stage takes very less time. This algorithm is meant to work for large datasets with over a million data points. In the future, we can use it as a scanner tool to check the client’s database for such harmful content and mark it up for them. FAIRMod will increase the overall health of the internet and its accessibility. Our goal is to achieve 90% accuracy while having faster detection speed.