Automated Factchecking at PyData Conference 2017
On the 6th of May, Will (our Director) and Mevan (Digital Products and Supporter Communications Manager) spread the word about the future of factchecking at the PyData conference at Bloomberg.
Mevan and Will focused on the future of factchecking: automation. More claims are being made than ever before and they’re spreading faster and faster. We need to keep up, and that’s where automated factchecking comes in. You can read more about it here.
We got some great feedback!
Michael Skelly (our Campaign Technical Director) also ran a hackathon which drew in more than thirty volunteers for a whole afternoon.
Projects included helping us monitor more data sources, tweeting MPs with live factchecks, helping us match known claims better, investigating how claims cluster together and whether we can detect soundbites in their earliest infancy.
Our volunteers divided themselves into teams and set to work on our suggested projects, or their own ideas. Dave, Frank and Adam worked on matching claims to existing conclusions. They converted claims and Hansard archives (spoken word in PMQs) into word vectors which were then used to match the most relevant factchecks to a given claim. They then applied this system to all sentences spoken by MPs (using the Hansard dataset) to test it.
Eron started making a Twitter chatbot; Oliver, Ester, Dyaa and Miguel investigated new sound-bites that are on the rise; Jenna, Charlotte and Ben streamed politicians tweets, matching claims with conclusions as well as writing a script to convert a tweet into haystack XML format. And that’s just a snapshot of what was achieved.
It was great to get so many people on board with a new era of factchecking; bringing together a huge range of skills and interests and getting the word out there.