NewsDetectives

The website when you first enter
When you input an unreliable headline
When you input a reliable headline

Inspiration

We were inspired to make an application which checks news reliability since we noticed the ever increasing amount of misinformation in our digital world.

What it does

Simply input the news headline into the input box and press submit. The algorithm will then classify it as a reliable, or unreliable source.

How we built it

We used Tensorflow with the Keras API to build a model with word embeddings in 32 dimensional space. Then we used a dense layer to make the final model predictions Finally, we used flask to connect the model with the front end HTML and CSS code.

Challenges we ran into

The most challenging part in building the model was dealing with over fitting because of limited initial training data. This caused the model to perform in training, but not work as well in testing.

Accomplishments that we're proud of

We are most proud of using data augmentation to increase the size of our training set which resulted in the model having 97% accuracy on validation. Our team was also very organized, respectful of each other, and most importantly, we all had a really fun time!

What we learned

We learned a lot about flask as our team had nearly no experience with technology. We also learned how to integrate our machine learning model into a website!