The problem Server Police-An Anti Cyber Bullying Discord Bot solves

Cyberbullying is an enormous problem faced by people across all age groups on the web. Despite several measures being taken to curb cyberbullying, the efforts of cyber police usually go in vain. Cyberbullying manifests itself in different forms. Hacking into someone’s profile and leaking sensitive, personal content, being an impostor, spreading false information that could defame a person or a group, introducing bugs and malware to gain control of the person’s device and use that further to get their way. A common type of cyberbullying that children fall prey to is the use of offensive language and posting negative comments with the aim of shattering the victim’s image. Consequences of such acts include harm to both the physical and psychological well being of the victim.

Our aim is to develop a bot that detects the use of such offensive language and passing of derogatory comments. Depending on the extent of the language used, the application on which the bot is deployed will take suitable actions against the offender. With the help of Deep Learning algorithms and open source libraries for the same, the bot will be built so as to be easily deployed on multiple platforms. In this prototype we'll be implementing our model on discord, using discordjs. There will be a bot in the server which can read all incoming messages and take actions accordingly.

Challenges we ran into

The data sets that were available online were not very accurate so we made our own data set and trained our model on it to increase the accuracy from 75% to 88%.

The model we built to detect the offensive language was in Python whereas we had to create the Discord bot on Node.js. We faced a bit of a challenge in integrating these as we were unsure if Python code could be run on Node.js. However, after a bit of searching online, we found an article on GeeksforGeeks which solved our problem. Python scripts can be run on Node.js using the child_process.spawn() method. We used this method and were able to integrate our modules successfully.