B for the BBC
Average veracity scores by category on BBC
Some member's were motivated by Black Pepper's challenge to identify Fake News, others wanted to try to use majestic's API, another simply wanted to give natural language processing a go.
What it does
FaktNews user with convenient dashboard of metrics that scrutinise a website, for it's association with other sites using the Majestic API, for it's content using NLP and sentiment analysis and for it's standing with the community (as represented by a thumbs-up/ thumbs-down feature).
How we built it
We wrote Python scripts that run from a Flask application, sending information to a chrome extension.
Challenges we ran into
It's hard to produce meaningful analysis on content, assigning truth values to articles is problematic and even more so for sites. Getting lightweight and appropriate JQuery plugins to display all the information we were generating.
Accomplishments that we're proud of
Collaborating well enough to have a working prototype. Familiarising ourselves with more libraries and better ways to do things.
What we learned
This is a massively complicated problem to tackle, interestingly there is very often very little substantive difference between a reliable and unreliable article. Ideally a program could independently fact-check, corroborating claims made by sites with substantive authorities. The Issue there would be recognising those substantive, ideally infallible sources.
What's next for FaktNews
Probably sell it for $1,000,000,000