Fleetometer - A News Article Scoring System

by sc999 » Fri Oct 05, 2018 2:03 pm

What's the motivation?

Here at fleetunderground.com we aim to be forward looking and innovative. To see where technological progression and new thinking and can improve services in the news space.

Looking at the phenomenon of fake news, we could see that the deliberate fabrication of facts was only part of a broader problem and that there were many other factors that could be tackled. This lead us to the decision to build the Fleetomoter, the world's first user facing AI driven news scoring system.

What is it?

The Fleetomoter is an automated tool making use of AI technology to calculate a quality score for each news article. Along with the overall score, we also report on four sub-scores as follows:

• Fleetomoter – Overall Quality Score
• Spinomoter – Measures Spin
• Biasometer – Measures Bias
• Fakeometer – Measures Fakeness
• Speculometer – Measures Speculation

Scores are presented as percentages, with 100% being the best, obviously!

How does it work?

The system works by analysing a very large library of articles which have each been labeled to either posses or not possess various contributing indicators to those in question. Using mathematical functions, models are created that on this scale can usefully separate the noise from the signal, to determine linguistic patterns associated with each attribute.

What's it for?

We believe that this sort of technology can be a genuinely useful tool for news readers in their discovery and consumption of content, aiding a healthy lens of skepticism perhaps, and will only improve and become more widespread.

In the future?

The model updates and improves over time, and it's self learning nature will be updated to harness user feedback, so that user reaction is directly incorporated back into the model and the score.

What are the limitations?

Ok, a word of caution. It won't be perfect at this stage. It's just an algorithm so add a pinch of salt and don't take it too seriously. This tackles a difficult problem, not only because of the complexity of the technologies involved and determining an effective recipe, but also because of the inherent difficulty of finding scores that different people will agree with.

Thanks, the Fleetunderground team

Like, dislike or are 50% on the idea? Let us know in the comments.