iNews is a project made for the Expo-Sciences 2020. It uses stance detection to guide the user into detecting fake and real news.
I created a deep learning model for stance detection achieving better performance than state of the art by applying Word2vec embeddings and multi-channel convolutional architecture and it was able to have a general accuracy of 85% compared to 82%.
Furthermore, I used Python and BeautifulSoup to scrape snopes and mediabiasfactcheck, websites that debunked fake news article and that verified the credibility of the source to display it on the dashboard. I also used Google News API to find similar articles to the one given and performed stance detection, showing a chart that described the proportion of articles who agreed or disagreed with the given article.
The website was made using React and the backend was made using Flask
Unfortunately, I decided to not deploy it given the cost for such compute. However, you can run it locally by following the steps below.
- Execute setup for
webfolder - Execute setup for
backendfolder