Twitter, as part of ongoing efforts to improve the “health” of discussions on the platform, announced that it has acquired U.K.-based artificial-intelligence startup Fabula AI.
Terms of the deal weren’t disclosed, according to an account in Variety. The initial focus for Fabula as part of Twitter “will be to improve the health of the conversation, with expanding applications to stop spam and abuse and other strategic priorities in the future,” according to Twitter chief technology officer Parag Agrawal, who announced the acquisition in a blog post on June 3.
Fabula has developed the ability to analyze “very large and complex data sets” to detect network manipulation and can identify patterns that other machine-learning techniques can’t, according to Agrawal. The startup has created a “truth-risk scoring platform” to identify misinformation, using data from sources including Snopes and PolitiFact.
Twitter, in addition to improving detection of spam and other violations of its policies, plans to use Fabula’s technology to enhance products, including the timeline, recommendations and the explore tab, as well as the process for how users sign up for the service. Spam and bogus accounts continue to be a big problem for the social platform. According to Twitter’s estimates, in the first quarter of 2019, fake and/or spam accounts represented fewer than 5% of its active user base.
Fabula’s team will join the Twitter Cortex machine-learning team. Twitter said it has created a research group led by Sandeep Pandey, head of machine learning/AI engineering, to focus on such areas as natural-language processing, reinforcement learning, machine-learning ethics, recommendation systems, and graph deep learning.
Founded in April 2018, Fabula is led chief scientist Michael Bronstein and chief technologist Federico Monti, who began collaborating together while at Switzerland’s University of Lugano. Bronstein is currently the chair in machine learning and pattern recognition at Imperial College London and will remain in that position while leading graph deep learning research at Twitter.
Twitter in the past has acquired other AI startups, including image-search specialist Madbits in 2014, machine-learning configuration developer Whetlab in 2015 and visual-processing startup Magic Pony in 2016.
The acquisition will underpin a research group at Twitter led by Pandey that will work toward finding new ways to leverage machine learning across natural language processing (NLP), recommendations systems, reinforcement learning, and graph deep learning. The group will also address ML ethics, according to an account in VentureBeat.
Fake News Spreads Faster than Real News
“Fake news” has become an umbrella buzzword to describe the deliberate spread of misinformation, but Fabula AI is really about helping identify the authenticity of any information that circulates on social media — regardless of intent. Studies have shown that false news spreads faster than real news online, a pattern that can be used to help spot misinformation. This is what Fabula focuses on: detecting differences in how content is spreading on social media and allocating an authenticity score.
“As this technology detects the spread pattern, it is language and locale independent; in fact, it can be used even when the content is encrypted,” the company says on its homepage. “We also believe that such an approach, given it is based on the propagation pattern through huge social networks, is far more resilient to adversarial attacks.”
As with most of the major social media platforms, Twitter has faced its share of criticism for the way it is used to spread misinformation. This latest move is designed to “improve the health of the conversation” on Twitter, according to CTO Parag Agrawal, and will expand over time to help stop other forms of spam and platform abuse.
“By studying and understanding the Twitter graph, comprised of the millions of tweets, retweets, and likes shared on Twitter every day, we will be able to improve the health of the conversation, as well as products, including the timeline, recommendations, the Explore tab, and the onboarding experience,” Agrawal said.
Last year, Facebook snapped up Bloomsbury AI, a startup building NLP smarts that could also be used to help combat fake news.
With the 2020 U.S. presidential election on the horizon, social media firms will be under intense scrutiny for their handling of fake news — which is partly why Twitter is looking to invest in automation to weed out the bad eggs.
“Machine learning plays a key role in powering Twitter and our purpose of serving the public conversation,” Agrawal said.