Utopia Analytics reduces volume of online abuse
Hate speech, trolling and misconduct are a constant, and growing, problem online. Until now. A new Finnish artificial intelligence-powered moderation tool gets rid of abusive messages in seconds.
It was by complete coincidence Mari-Sanna Paukkeri that became interested in text mining, the science of deriving quality information from large amounts of text. She heard about the topic while studying physics at Finland’s Aalto University and has not looked back since.
“I went to an orientation session where they explained the amazing things you can do with mathematical algorithms and I was sold,” she recalls. “I changed my major there and then.”
Skip read to today and Paukkeri is not only behind the world’s first PhD dissertation on language independent text mining, but the CEO and co-founder of Finnish artificial intelligence (AI) and analytics startup Utopia Analytics. The 2014-founded company uses unique machine learning algorithms to manage, analyse and visualise large masses of text in any language.
The company’s latest product, Utopia Moderator, applies this technology to tackle hate speech and misconduct online. Whereas human moderators need breaks and, no matter how capable, are limited with the amount of data they can process, Utopia’s AI-powered moderator keeps going 24/7. The software learns from examples how the moderation guidelines of a company work and swiftly removes any inappropriate messages, leaving only the most difficult cases for its human colleagues.
Easily understandable language
Utopia Analytics’ moderator tool is opening up a completely new market. While spam filters and basic moderator tools already exist, they lack the language and learning capabilities of the Utopia Moderator. The patent pending technology is capable of understanding any of the globe’s languages and is not thrown off by typos or slang.
“Language is used very carelessly on social media, but that does not matter to us, as we can moderate any writing style,” Paukkeri says. “We have compared our product to different spam filters and they can moderate 20-30 per cent of all hate speech with a high error rate, while we can moderate it all with high accuracy. It is a huge difference.”
The attitude was nothing can be done about hate speech online and now we have a tool for getting rid of it.
In a pilot project with one of Finland’s largest chat forums, Suomi24, Utopia Moderator helped to keep check on over 16 000 messages posted to the site daily. During the project the AI and human moderators agreed 90 per cent of the time.
But the competition is heating up. In February, tech giant Google introduced its own moderator tool (Perspective API), but for Paukkeri this is positive news.
“Its features are very bare compared to our product and it only works in three languages,” she explains. “In a way [the Google product] plays to our advantage. They can easily get publicity for the technology and people start to be aware that this is possible, which makes our sales efforts easier.”
Talk of the towns
Currently, Utopia Analytics has customers in Finland and the Baltic countries. In addition to publishing houses, the AI moderator helps dating sites and online marketplaces detect illegal ads and misconduct.
While the startup is now focused on expanding its international reach, its team of 10 employees (four of which have a PhD in AI or text mining) is also working on introducing new products and features. The first will add image moderation to its offering.
Paukkeri says the technology can be used for any industry where large masses of text need analysing, from customer feedback to legal documents.
“We see there are huge markets where we can help people to do their jobs,” Paukkeri says. “Already with our moderator product we have turned one thing upside down. The attitude was nothing can be done about hate speech online and now we have a tool for getting rid of it.”
Text: Eeva Haaramo