FetchTruth

Logo Created with Figma

Problem

News is a tool of communication, but what happens if disinformation is more prevalent than accurately sourced news? Disinformation, or “Fake News,” can negatively alter the public's opinion of important topics that affect our everyday lives. Politics, climate change, and identity are some of the biggest victims of disinformation and can inflame heightened tensions worldwide. We created FetchTruth to assist with critical thinking when it comes to reading news sources.

Challenges we ran into

Wanting to incorporate user data into the application and the actual data analysis. We wanted to know whether we should collect user data without having to log in, as of now, and we decided to accept user responses for our interpretation of whether a news source is valid or not and their trust level. We will save all of this for future data collection.

How we built it

Called Huggingface fake news discriminator LLM to distinguish whether a news site can be trusted based on a model. We carefully selected a well-known and reliable model from HuggingFace, and one that had a large dataset that could be properly cross-referenced. The responses will also be added to another database, which then have analysis performed on it.

Huggingface Model

Future Directions for FetchTruth

Integrate user feedback into model to make decisions for specific and more categories, like why people trust sources with specific keywords, etc.
Create user profiles and connect visited articles among users
Tagging system for the most popular topics and keywords for better lookup for users
Scraping all data on news article to do a deeper analysis of articles