Fact Checker Chrome Extension

Fact checks highlighted or input information in real time using many LLM models; determines an inaccuracy score and provides detailed insight for that score.

Comment

Inspiration

We were inspired by the desire to rid the world of misinformation being spread by many malicious news outlets and other forms of media. We also want to help older generation of people who are more susceptible to misinformation.

What it does

Our project is a web extension to chrome browsers that enable people to access our multi-AI fact checking service. It lets the user quickly check any text they highlight in chrome webpages as well as input their own statements into the pop-up to be checked for truthfulness. Furthermore, it can provide deep insight when it comes to the reasoning that explains what we call the inaccuracy score. We also provide a natural text-to-speech service using Eleven Labs for the elderly that might not be comfortable reading smaller letters on a screen.

How we built it

We built this extension using a mix JavaScript, CSS, HTML (for the front-end necessary for chrome extensions) as well as Python and Flask (for the back-end and front-end-to-back-end-communication). We (almost successfully) attempted to use Vultr to host our back-end processes publicly and Eleven Labs to implement natural text-to-speech.

Challenges we ran into

Mainly, we struggled with implementing our back-end into a cloud to make it publicly accessible due to a variety of issues, but we ran into some issue (seemingly for authentication) which we couldn't overcome. (We did get the the CPU running and were able to host our back-end there, but the communication was ruptured somewhere). Also, faulty API keys greatly delayed our project and prevented video footage. Furthermore, we had our fair share of issues with GitHub which made documenting and version control a bit more of a challenge.

Accomplishments that we're proud of

We are proud of our simple, intuitive and accessible UI and of our solid Python back-end able to handle communication with 3 APIs, a cloud server and a JavaScript front-end while being relatively robust. Finally, we think our solution to the misinformation problem is quite adequate due to its accessibility, ease of use, and suitable quality.

What we learned

We learned that dealing APIs can be arduous; we also learned that even simple things like chrome extensions can be used for quite complex and useful tasks. We also learned to use Flask, JavaScript, and improved our handle of git.

What's next for Fact Checker Chrome Extension

We plan to implement more cross-LLM verification in order to massively improve the accuracy of our models' analysis. Furthermore, we know the UX can use some improvements, particularly when it comes to the speed of our analysis; optimization is a must!

Built With

Share this project:

Updates