What is fake news? That is a difficult question to answer. Is it satire? Is it opinions pieces? Is it factual inaccuracies?
With the proliferation of social sharing and user submitted content, we’ve opened up our democracies and institutions to a level of abuse, rhetoric, and influence that is unprecedented.
Fortunately, we think machine learning can help.
The fake news problem might be able to be broken down into a text classification problem. If you simplify the problem, you can create two or more classes of news, and find examples of each class, then train a machine learning model to understand the difference.
Machine Box has a tool called Classificationbox which lets you easily create classifiers that you can run in production. We also have a nifty open source tool that lets you simplify the training process even further with text files called textclass .
Using these two tools, you can do a great number of things. Your challenge is to classify news in a way that will benefit your users. Perhaps they only want to read about sports or they’d like to have articles that are opinionated ranked higher on a list of new articles for the day.
The following are some great tutorials you can follow to get started:
- I trained fake news detection AI with >95% accuracy, and almost went crazy
- How to train a spam detector with a 97% accuracy with Machine Box
- How I trained an AI to detect satire in under an hour
- Build your own fake news detector using machine learning
Whatever you end up building, whether it is a challenge you’re taking on yourself, or as part of a hackathon, please share your results with us. We love it when you build amazing things with Machine Box!