We want to protect the world from fake and real news media and stop the witch hunt.
What it does
It protects the world from fake and real news media and stops the witch hunt.
Provides an endpoint to get a score for fake news and real news. We also have a little UI for that.
How we built it
We trained a deep learning (MLP, KNN, LR, Ensemble, NLP, TF-IDF, Voting) model with labeled data. It's fake news that we used a random forest in the beginning. This model is serving predictions via a REST API to our front-end service.
Challenges we ran into
- stay focused
- have enough drinks (stay hydrated)
- RAM (Random Access Memory)
- Suomi fighting against SY (totally random encounter between us and them)
Accomplishments that we're proud of
- we have a working prototype
- we even implemented grid search!
- we have a front-end without having front-end developers in the team!
What we learned
- the SRF database is only available in Italian
- the SRF data is not labeled (We assume that they assume that they only serve real news)
What's next for Trump Detector
- use better training data
- improve the front-end (but that will be tough to beat)
- 45's approval
- The Orange One's approval
- Android / Windows CE and Cisco iOS support