Suggestions that AI makes are really helpful for targeted news, which helps companies like Google and Facebook to improve their models to give targeted information according to the likes and dislikes of users.

Although suggestions lead to Targeted news, targeted news can lead to heavy Manipulation! This Manipulation can be used as a Weapon by 3rd party countries to change people's minds...which can also result in war!

So to solve this massive problem, we decided to build an application to help protect people from getting false information

What it does

What does it do? Step 1: User Enters URL of the site to be verified Step 2: The website will be scraped and the information will be converted into a text file Step 3: We use our Robust Self Trained Model to predict if the news is fake or not

How we built it

We Built and Trained a Machine Learning Model from Scratch and Used Natural Language Processing to process our data. To go into a little more detail, we used: ->Techniques such as Lemmatization and CountVectorizer, ->Modles such as Bag of words model ->and Algorithms such as Multinomial Naive Bayes Algorithm and Passive-Aggressive Classifier Algorithm.

We used our Own Data Sources (of more than 65k rows!): they consist of data from the University of Victoria fake news and real news data set, and A Kaggle fake news and real news data set.

We designed the application for consumers and used Figma to launch a prototype due to the lack of time.

Challenges we ran into

Some of the Challenges we ran into were: 1) While merging the data sets, we needed to make sure we knew what columns we were dropping to have a good output to the model 2) While testing the data with other articles, we figured that our shape of the vector we were feeding into the model didn't match one of the input models. But with the help of some godly magic, we were able to get over with it pretty easily :P

Accomplishments that we're proud of

1) We are proud that our model reached a 90% Recall and 85% Accuracy! Which we felt was pretty impressive 2) We are really happy that we are solving one of the biggest problems the world is going to face soon and we think that this will be a stepping stone to some mandatory precautions we will have to take in the coming years

What we learned

We learned a lot from this hackathon and we will take away a lot from this.

What's next for Fake News Detector

There is a lot more to Fake News Detector and we will work hard to make more improvements to it!

Built With

Share this project: