We were inspired to create this project due to the avalanche of recent fake news and conspiracy theories relating to vaccines and the 2020 election. Misinformation spread through social media and untrustworthy news sites have caused confusion, led to violence, and have negatively impacted the health's of millions of Americans. We were determined, therefore, to do something to solve it.
What it does
It uses a machine learning model trained on thousands of news articles to detect if a news article is fake or not based on its contents. It first takes the text and then converts the words to vectors using WordNet. This is then passed to the model which outputs 0 for fake and 1 for real.
How we built it
We built this by using a publicly available dataset (found on kaggle) of news articles labeled as REAL or FAKE. We trained the model on a subset of this dataset, and then evaluated it on a test set on which it got 90+% accuracy. We loaded the data with Pandas and made and trained the model with sklearn and Google Colab GPU.
Challenges we ran into
We had problems with deciding a dataset, as we needed to find one that had a variety of data and would not be limited to a specific area of news (i.e. just politics or just science). We eventually found a good dataset by going on Kaggle and comparing each datasets diversity as well as the overall performance of models trained on that dataset.
Accomplishments that we're proud of
We're proud of our model's high accuracy of 90+%. We believe this model will be able to have numerous applications in social media sites like Twitter that aim at stomping out fake news or search engines like Google that can alert a user if a searched site appears to be pushing fake news.
What we learned
We learned about the prevalence of fake news in people's lives and the negative effects it can have, and we also learned how to apply sklearn and machine learning for language processing and classification.
What's next for FakeNews Detector
We are hoping to create a website or a chrome extension for FakeNews Detector that can tell you in real time if a website you are viewing is fake news or not.
Log in or sign up for Devpost to join the conversation.