We all have faced cyberbullying in our lives at some point or the other. It is one thing to joke around with your friends, and another to be hurtfully called out on the internet. Such remarks are made with a negative intend. These trolls choose to get their few chuckles by dragging down other people and harming them. While such cases rarely involve physical interactions, these lead to deeo trauma and cause severe impact on the victim's mental health. At times, these victims fall into depression, body dysmorphia, social insecurity, etc. Cyberbullying is the modern day monster that we are all ignoring.
Another issue that we face is the misuse of language diversity. Natives would often throw remarks at people who are unfamiliar with a language or are not native speakers. Unfortunately, this is now translating to online social media. Scrolling through comments, it is common to come across racist remarks. Moreover, these are at times hidden in the context. That is, someone who is not very comfortable with the English language would be unable to figure out what the words would mean. This just widens the language barrier and hampers inclusion.
"Cyber-bully-detector" is an extension or a plugin which intends to scan through the text and highlight any such remarks on the web.
As of now, the project is still in its planning stage but we have thought about the following as the probable path:
A Chrome Extension to gain access to client side data and relay back the processed data;
A RestAPI as a pathway for exchange of data;
A PyTorch model to classify the text into derrogatory or otherwise;
The NLTK library for sentiment analysis.
While we are focusing on the implementation in English language for now, we intend to expand to several other languages as well.