Website for Online Trolls Detector using Deep Learning

Internet trolls seek to create chaos on social media spreading misinformation. For this issue, I worked on developing a deep learning model that detects if post/comment is from a troll or not.

Comment

Discord: selvera#7881

Inspiration

There are many internet trolls on social media spreading misinformation and causing chaos. This projects aimed at detecting these trolls by analyzing their post/comments. These comments come in the form of sentences(string), so I used Natural Language Processing deep learning state of the art model BERT by Google.

What it does

Two features were developed: website and deep learning model. The website purpose was to allow users enter a comment that they thought looked like a trolls comment. The website would use the deep learning model to classify if it looks like a troll comment or not. The deep learning model was trained on reddit labeled data to learn the structure and distribution of trolls comment.

How I built it

I used Heroku and Flask to build the website. For the deep learning model, I used pytorch(popular for Natural Language Processing) library for developing and training the model on google colab (free gpu).

Challenges I ran into

I have never developed a website, so I spent most of the time on that. A lot of tools to learn and use. The biggest challenge is still to use deep learning model for the website. The model expects a gpu to make the classification. But of course the web host will not provide a GPU. Therefore, I still have to learn how to adapt a model that uses GPU to be deployed on free web host.

Accomplishments that I'm proud of

I am proud of having a website running. Now, I can showcase my future work. Also, I am proud of creating a model with high accuracy.

What I learned

I learned how to develop a website and the whole process of get it running. I also learned that deploying a deep learning model on a website is a hard task and that actually is new kind of problem being tackled in academia and industry now.

What's next for Website for Online Trolls Detector using Deep Learning

Implementing the website with the built model to allow users enter the comment and provide them feedback.

Built With

Share this project:

Updates