• Aucun résultat trouvé

17043 Google funds automated fact-checking software in bid to fight fake news

N/A
N/A
Protected

Academic year: 2022

Partager "17043 Google funds automated fact-checking software in bid to fight fake news"

Copied!
1
0
0

Texte intégral

(1)

17043 Google funds automated fact- checking software in bid to fight fake news

Project aims to develop software which can automatically fact-check news 'in real time'.

The Independent – Caroline Mortimer – Thursday 17 November 2016

Google has agreed to fund a project to develop automated fact-checking tools amid anger over the prevalence of fake news websites during the US presidential election.

UK fact-checking organisation, FullFact, has announced it has been awarded €50,000 (£43,000) by the tech giant’s Digital News Initiative to build the first “fully automated end-to- end fact checking system”.

In a statement, FullFact explained that the system will have two main features. One will inform readers if something reported as fact has already been proven inaccurate.The other mode will fact-check claims automatically using Natural Language Processing and statistical analysis in real-time by highlighting the text and having a factbox appear when the user hovers over it.

The project, which is also supported by two smaller UK tech companies Bytemark and Flax, aims to fight fake “news” sites disseminating false information – particularly about politically charged topics such as perceived crime by refugees, immigration and the alleged crimes of politicians such as Hillary Clinton and Donald Trump.

According to analysis by Buzzfeed News, fake election news got more attention on Facebook than real stories during the US presidential election.

It found that the 20 top-performing stories from hoax websites and partisan blogs generated 8.7m shares, reactions and comments in the three months to polling day – compared to 7.4m for conventional news outlets.

Facebook attempted to diffuse criticism saying it was a tech company not a media company and was not responsible for the information shared on its network.

But following the shock victory of Mr Trump last week, an unofficial task force of Facebook has reportedly been set up to examine the role it played in the spread of false information during the campaign.

After a report found Facebook’s News Feed feature was skewed in favour of left-wing news sources in August, the company fired some of its "trending news" editors but the algorithm which replaced them automatically began promoting false news shared by users.

FullFact said the initiative was about restoring faith in politicians and public figures following the EU referendum where many were perceived to have openly lied.

It said: “A lot of people on both sides felt lied to during the referendum. What's worse is that many people think it's inevitable they will be lied to. We all need tools that help us choose when to trust and when not to trust claims we hear. Otherwise, it’s too easy just to switch off completely and be cynical about everything.”

410 words

Références

Documents relatifs

We consider our analysis of fact checking as a content management problem to be new and original, and we are eager to present the many opportunity for research raised by data

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

•  AutomaDcally find most relevant reference data for a claim. –  RDF dump of full

Following this, I will present our approach for fact checking simple numerical statements which we were able to learn without explicitly labelled data. Then I will describe how

Our approach for this task was based on the hypothesis that factual claims have been discussed and mentioned in online news agen- cies.. In our approach (Ghanem et al., 2018a), we

The system then aggregates this body of evidence to predict the veracity of the claim, showing the user pre- cisely which information is being used and the various sources of

Why would they trust gov.uk to state what is the case about a random company? If that sounds implausible, think of it the other way around: why would a

The key idea behind our work, which constitutes its main novelty, is that hoaxes can be identified with great accuracy on the basis of the users that interact with them. In