Rosette brings the text analytics power to news data startup, Zetta Cloud

The US presidential election has made it clear that fake news is spreading across the internet like a virus. While shared knowledge is one of the many perks of our increasingly connected online society, it also has a darker side: the opportunity for misinformation. Romanian news data startup, Zetta Cloud, seeks to bring trust and veracity to the online news community through their application TrustServista. Loosely translated, “servista” means “butler” in Esperanto and TrustServista serves as your personal news verification assistant. The tool provides you with knowledge to decide whether or not to trust news you read online, before you even ask for it.

In a traditional newspaper setting, journalists vet their sources to ensure that the information they’re sharing is true, accurate, and unbiased. However, millions of biased news sites, clickbait articles, and satirical pieces have little or no source verification process in place. Some sites are purposely misleading or spread patently false stories designed to maliciously “troll” unsuspecting readers. After all, this is the internet and anyone can distribute anything. However, if these false articles are accidentally picked up or contaminate a source used by a major newspaper or network, the ramifications can be disastrous. News outlets have a reputation to maintain, and their readers’ loyalty depends on it.

These days, news distributors find themselves in a catch-22 situation: they want to preserve their integrity by reporting accurate news, but, they’re under pressure to break the “big story” before their competitors. Furthermore, the time needed to trace a piece of information all the way back to its source can be impractical or exhausting. What reporters today really need is a tool to automate the vetting process in real time, so they can safely leverage vast data lakes of internet news without risking their reputation. This is where TrustServista comes in.

TrustServista in action

TrustServista is a machine learning-based tool that calculates trustworthiness scores for online articles. Its dashboard allows users to easily track where the article’s information is coming from and how individual trustworthiness scores were determined.

When the team at Zetta Cloud first started building TrustServista, their initial findings surprised them. “One article from Reuters has a half-dozen links and references that each have another half-dozen more related posts.” says Emil Stetco, co-founder and CEO. “A simple story can have hundreds of sources built into it. To know if an article is trustworthy you have to check every source article and trace it back to ‘patient zero.’”—referencing the initial case of an epidemic.

In the field: protecting journalistic integrity

In late September 2015—just weeks after the Volkswagen emissions scandal roared through the auto industry—a German automotive magazine Auto Bild published an article, “AUTO BILD exclusive: BMW diesel also clearly exceeds exhaust gas limits.”

BMW stocks immediately fell over 9% as the news spread from blogs to news sites to social media posts. By the next day, Auto Bild had retracted their statement and clarified that, unlike VW, BMW hadn’t been cheating; their diesel engines just didn’t perform well on one of the ICCT tests. Unfortunately the damage had been done.

Zetta Cloud CSO, George Bara said, “It’s like catching a moving bullet. Some of these articles can go viral on the web in less than one day.” A tool like TrustServista could have prevented or limited the proliferation of the faulty story, saving both reputation and money.

In the field: being socially responsible by scoring social media

Fake and misleading news stories can have social and political ramifications as well. “Social media has taken on the role of a major distributor of news content.” says Bara. “There is a responsibility that comes with that.”

Following the tremendous volume of fake political news stories propagated during the recent U.S. election, social media giants have been facing enormous pressure to crack down. Facebook released a plan to utilize user reporting and fact checkers to limit false news, but this is still only a reactive solution. Although, Facebook could attempt to automatically delete posts they deem un-newsworthy, they would run the risk of violating user trust and free speech protections, as we saw recently with Reddit CEO Steve Huffman.

TrustServista, however, offers a reasonable solution. By providing a trustworthiness score with all news stories, TrustServista equips users with the information to judge whether to read or trust an article. “We will also transparently show you WHY that article has that score.” adds Stetco. The user can choose to proceed regardless of the score, but TrustServista has done their due diligence—warning users of potentially untrustworthy content while preserving an open online community.

Behind the scenes

The Zetta Cloud team, led by Stetco, Bara, and CCO Sebastian Ionita, is tuning a proprietary “trust” algorithm that factors in a number of different patterns they’ve discovered through the process of analyzing reputable and not-so-reputable news data.

“We utilize sentiment analysis to recognize clickbait articles,” says Bara. “Clickbait works by appealing to sentiment—causing strong emotion in people. A proper piece of journalism should not have a lot of sentiment in it.”

Trustworthy news sources on the other hand frequently mention names and locations. They stick to the facts rather than extrapolating or spreading opinions. TrustServista relies on Rosette’s entity extraction to help identify the articles that fit the profile.

Given the international nature of the news, TrustServista also needs to understand a multitude of languages. If that “patient zero” article was in Greek, the tool needs to be able to analyze that original Greek text. Rosette, which provides multilingual entity extraction in over 20 languages, was the obvious choice.

Rosette for startups

Beyond the strong technical stack and language coverage, Zetta Cloud was drawn to Rosette’s affordable startup program and responsive support community. The team found other text analytics companies to be slow and bulky; it took nearly two months just to get pricing! By signing up for a free Rosette API account the Zetta Cloud team was able to quickly test functionality and work with the support team to address questions and issues.

“It’s very hard to find partners who have the same mindset as us,” says Bara. “It’s not easy to sell or to be sold to, so when I find companies who are agile and want to help us succeed, I’m excited to work with them.”

“It feels like you guys care about the customer.” Stetco continues. “You come back to us with solutions that are outside your platform if necessary. We said, ‘Whoa, that’s the support we’re looking for, and what we should offer our customers as well.’”

Making a difference

It’s unlikely that fake news will disappear anytime soon, making Zetta Cloud’s TrustServista application an important and necessary tool for today’s reporters and readers. Social media sources are taking over the role of traditional news distributors, but without the same transparency. TrustServista, with the help of Rosette, allows readers to understand why an article is trending and where it came from, not just how many people have clicked on it.

For more information about Zetta Cloud and TrustServista, visit their website, follow them on Twitter, or email contact@zettacloud.ro.