Increased cases of depression and raise in suicide rate caused by cyber bullying.
What it does
It recognizes and blocks not only abusive words but indirectly offensive phrases using the might of machine learning.
How we built it
We collected datasets from various sources and also created our own dataset for the training of our model. The Frontend part was created with React.js and was connected with our model using Flask.
Challenges we ran into
- Finding and processing imbalanced datasets
- Raising the accuracy above the previously researched level
Accomplishments that we're proud of
We achieved an accuracy of 93.75% which is much greater than previous research!
What we learned
- We learnt animations using React-pose library.
- Implementing and fine-tuning Recurrent Neural Networks.
- Connecting different languages and frameworks.
What's next for CyberCow
We are planning to further increase the accuracy and integrate it to various social media and online gaming platforms.