Invention:
The problem this invention solves is that of spurious patterns or invalid results from software neural networks tasked to use natural language programming for fact verification in news or articles. This approach increases the accuracy of fact verification. Neural architectures with improved transfer between domains enables re-use of trained models, saving protracted re-training periods. This enables ranking of content, such as news articles, based on their factuality.
Background:
We’ve witnessed growing political upheaval in the last few years, thanks to the proliferation of “fake news” online — much of which is disinformation that’s deliberately designed to spread virally and mislead public opinion for political purposes. Research shows that on average, a false story will disseminate to 1,500 people six times more quickly than a factual one, especially if it’s related to politics, and particularly when bots are used to further automate the propagation of fake news. Enormous amounts of content placed online every day render human fact checking impossible for the magnitude of fake news and false information. Tools utilizing deep learning, machine learning, AI, and algorithms to fact check speeches, posts, and articles are increasingly needed. While cable news networks and print publications may have human fact checking resources, most online information is not automatically analyzed by computer systems. Having the option of access to AI fact checking via a web browser extension would improve access by the general public to highly effective verification. An extension with this capability would be able to alert users if a site provides potentially spurious information/fake news.
Applications:
- Speech fact checking
- Media sources fact checking
- Browser extension
- Business software consultant agencies
Advantages:
- Improved accuracy
- Less spurious patterns
