6 Dec 2017

Excellent marriage between machine and human creates ‘Reuters Tracer’ lovechild

SECTOR: News

CLIENT: Thomson Reuters

WOW: Social Media News Tracer

DATE: October 2017

 

Las Vegas shooting, October 2017, 1:22am, 58 people dead. Earliest report on Twitter happens simultaneous to the shooting.


Every news organisation has an acute problem with fake news distorting the perception of events and there is an added pressure to break news stories as they happen.

The new system called Reuters Tracer examines 12 million tweets a day (2% of the daily total), validates news events and assigns a newsworthiness score with a confidence rating on how likely the events are true.

The algorithm  uses data mining and machine learning to pick out the most relevant events, determine the topic, rank the priority, write a headline and a summary. The human involvement comes from a list of Twitter accounts, curated by Reuters journalists.

A citizen journalist tweet at 1:22am about the shooting triggered what is called a ‘Tracer Cluster’ and, after rigorous criteria has been met, the event was included in the news feed at 1:39am

What we love most about the Tracer product is the cooperation between serious data brains in R&D and the human expertise brought in from the Reuters journalists – it’s a lot like how we work here at Mediacells!