Inspiration
I was inspired to build this project after seeing the spreading of fake news all throughout the world in all different languages. This website helps citizens take charge of their news and determine what is real versus fake which helps keep them informed and avoid falling prey to propoganda. This is important as information warfare has become increasingly more common and convincing in the 21st century.
What it does
This application allows you to paste a news article, select the language, translate to English (if needed), and then analyzes whether the news is legitimate or fake.
How we built it
I built it using streamlit to host the page and allow user input which uses python. The data science portion was also done in python as well. For the model I trained, I used NLTK for preprocessing and sklearn for the test/train split and models.
For the news analysis I ran 7 different models that had an average accuracy of 94.15789532754575% with some being as high as 97%. I then put each model's decision in an array and took the median of the array. This helps counter any algorithms that may return a faulty result and ensure that the correct result is delivered.
Challenges we ran into
Challenges I ran into was not having enough domain names to register, issues with a last minute deployment to streamlit, and homework for other classes.
Accomplishments that we're proud of
I am proud that I worked alone on this project and was still able to complete it in time as a finished project. I would have liked to have trained all the language models myself, but I am happy that I was able to get this done as a good first draft.
What we learned
I learned about streamlit apps and more about the Sklearn library than I had ever learned before. I also learned how to deploy a website which was something I had never done before.
What's next for FakeNewsDetector.AI
Next for FND.AI is more languages to test. This will allow for even more fake news to be exposed. This could help expose any campaigns against smaller communities' languages. I also want to deploy to streamlit and my own domain name too.
Log in or sign up for Devpost to join the conversation.