FML Blog

In touch with innovation: talking to Zetta Cloud

Tuesday, September 12, 2017      Future Media Lab. team       0

The Future Media Lab. team is continuing our “In touch with innovation” series on our blog with a series of Q&A’s that were conducted with media start-ups from across Europe. In this edition, we spoke with George Bara, the Co-Founder and Chief Strategy Officer at Zetta Cloud in Romania. ZettaCloud provides digital investigation tool - called Trust Servista that automatically determines the origin and trustworthiness of online news stories called Trust Servista. 

 

Future Media Lab.: What is the main idea behind your company and your product?

George Bara: Zetta Cloud was founded in 2013 with a direct focus on open data analytics and digital news software solutions. Our flagship product - TrustServista - was launched in June 2017 as a response to the fake news proliferation phenomena that affected the US and France presidential elections. With a mission to automate a lot of the work that is performed by human fact checkers, media journalists and intelligence analysts, TrustServista uses Artificial Intelligence (AI) to automatically determine the origin and trustworthiness of online information so that publishers and content distributors can take immediate action, as events unfold, rather than after the misinformation has spread out and reached beyond control. Currently, TrustServista processes more than 8000 articles a day in English and starting this week we also launched the German instance of TrustServista that will support DACH region news publishers and news agencies.

 

FML.: What sparked the idea of a news verification report?

GB:One of the key aspects of calibrating and improving our algorithms is constantly benchmarking the content that TrustServista captures. This is how we create internally statistical “clickbait recipes” or determine how to pick the best article to pick as “Patient Zero” or P0 (the original source of information) for a given story, metrics that is part of TrustServista. We realized that the wealth of information we are generating with these benchmarks can have value for the industry and the public, as it contains numerous insights into how AI algorithms work with real-life news articles.

 

The report also shows how the media landscape is structured and how the readers’ trust in the different content providers can be measured by looking at both how the content is created but also on how it is being distributed on social media. While most solutions for tracking “content performance” mainly looks at social media metrics and website statistics, we, rather look from a higher-level, comparing how different types of content providers (news agencies, newspapers, magazines, blogs, aso) differ in terms of content quality and user loyalty and draw conclusions into the actual trustworthiness of the entire English-language media landscape.

 

FML.: What are the main takeaways from the News Verification report?

GB: One key takeaway is that the established online media (newspapers, magazines, news agencies) still play the most important role in news topics generation. Both in terms of the actual volume of articles per day produced, in comparison to blogs for example, but also in terms of social media interactions.

 

We also found that negative topics are dominating mainstream news, while negative sentiments that are specific to clickbait techniques are dominating blogs. Both mainstream media and alternative media (blogs) are using the readers’ natural reaction to negative news in order to generate reactions, website traffic and social media interactions. The main difference is that while newspapers and magazines tend to have more “likes”, blogs will have more “shares”, showing that their loyal reader base has quite a different mentality towards their content.

 

In terms of sources used for creating content, social media platforms such as Twitter do not pay such an important role as people think: established news agencies and newspapers do not merely rely on Twitter for information gathering as much as blogs.

 

FML.: Is Artificial Intelligence (AI) able to perform news investigations?

GB: Yes, it is. The current Natural Language Processing algorithms can automate most of the tasks that human analysts perform. However, it still requires a trained investigator or analyst to give the final verdict of an article being true or fake. What AI can do is to save time. The time spent by teams of analysts or investigators can be reduced to a fraction using specialized tools.

The three main areas of investigation, understanding where the information originated, analyzing the text itself and verifying the credentials of the publishers, can be broke down into small tasks that can be handled by Natural Language Processing in combination with Machine Learning algorithms. Hoever, as with any AI-based technologies, the algorithms need calibration on actual real data (and a lot if it!), so there is a learning curve for the machine also. 

 

FML.: What is your added value to media companies in content curation/dissemination/monetization?

GB: Simply put, TrustServista saves time and gives clarity and overview to the storytelling process. Whenever information needs to be collected and matched in order to produce content, TrustServista finds all relevant content, structures it and offers a forensic-type approach to visualize it. And it does this in real-time, being able to handle even events that unfold rapidly and that generate content from different types of media, including user generated social content.

Even more, its capabilities to flag untrusted content can prevent editors from slipping in fake news or propaganda, thus producing high-quality content. And it does this automatically, with little manual involvement in collecting, organizing or analyzing the content.

 

FML.: What is your next big challenge/goal?

GB: Our main goal is to make TrustServista language-agnostic. The initial English-language support was complemented recently with support for German, Spanish, French and a couple of other languages. Having an analytics and verification engine that can trace information across multiple languages, performing multilingual search and verifying if a news originating from Russia (for example) has made its way to German media, would enable any media professional or analyst to get valuable insights into information even beyond what Google can offer, for example.

 

Here’s a link to TrustServista’s News Verification Report .

 

Editor’s note: The Future Media Lab. recently teamed up with next media accelerator to publish a compendiumof start-ups from across Europe who are working on solving some of the challenges facing the media sector today, particularly around content creation, monetization, dissemination and curation. Many of the start-ups were present to ‘pitch’ their solutions at the Future Media Lab. annual conference , which took place in Brussels on 2 May 2017. A digital version of the compendium is available for download (pdf) here.

 


Comments
Post has no comments.
Post a Comment

Captcha Image