TrustMeBro is a fake news and scam detection system that leverages GPT-4 to verify the credibility of tweets and TikTok videos. It was built for the Microsoft "Accelerate Your AI" Hackathon 2024 in Atlanta, where it was also awarded the third place.
- Social Media Post Extraction: Pulls tweets using the X api and can transcribe Tiktoks using the Google Cloud API
- Fine-Tuned Models: Two GPT-4 models were fine-tuned in order to achieve high accuracy when detecting the type of potential threat
- Keyword Extraction: Automatically extracts key claims from a tweet or video in order to ascertain their validity
- Wikipedia/Online Source Retrieval: Fetches summaries from Wikipedia or other online sources for relevant topics
- Credibility Analysis: Uses GPT-4 to classify the tweet as "credible" or "fake news" based on Wikipedia information
- Supporting Links: Provides source links if the tweet is flagged as potentially fake news
- TrustMeBro is hosted as a Streamlit app, accessible at trustmebro.streamlit.app
- You can also clone and install the repository on your computer. In order to do so, note that you will require the
openai,wikipedia-apiandgoogle-cloud-videointelligencePython packages
