evgenyboxer/fake-news

Using deep learning to train a model to distinguish between fake and real news

Fake News Classifier

Overview

This is an LSTM RNN model that classifies news stories as fake or not.
At this point, the answer is the probability of a given story to be fake.

The data for training this model is taken from: https://www.kaggle.com/clmentbisaillon/fake-and-real-news-dataset

Usage

import requests

url = 'http://localhost:9000/predict'
data = 'the news are TRUE, we all must believe that!'
r = requests.post(url, data)

print(r.json())

data is where the news story is stored (as python string). At this point the usage is local, the plan is to make a web page to serve the predictions.

TODO

1. Deploy to Heroku
2. Front-end work
   1. Make a web page with a text box to paste the news story and a predict button
   2. Once the model returns the str answer, it should be displayed in a designated place on the page (maybe an answer section)
3. Improve logging