Strategy for Bolstering Brand Safety Online Combines AI, Human Linguists
Brands like Procter & Gamble, AT&T, Netflix, Adidas, and PepsiCo. have all experienced it. The backlash that can occur when ads are placed alongside hateful online content is severe enough that some brands are pulling back on online spending and rethinking their digital media strategies.
Despite promises that they would do better, platforms like YouTube, Facebook, and others are still struggling with the issue. Brands don’t want their ads appearing alongside extremist content and hate speech, but flagging every piece of content that could be considered inappropriate is not an easy task.
The challenge has opened the door for a new industry of “authenticators,” which use technology to help brands avoid inappropriate content online. Using artificial intelligence and machine learning, these technology providers are usually able to evaluate the quality of an ad impression in real-time and help their brand clients avoid anything that could be considered inappropriate. Or at least, that’s what the goal is.
Brands are finding that AI is not a magic bullet for policing hateful content online. Although advancements in AI are happening every day, the tools that are available right now are not capable of determining the correct context and meaning of every word that’s uttered online.
If technology isn’t the solution, then what is? One idea that’s being floated is to combine AI with human linguists, who are more capable of understanding the meaning of words online and in different contexts. A firm called DoubleVerify, which authenticates the quality of digital media for brands, is championing the strategy and launching a dedicated “Semantic Science” team.
Language and content issues that seem simple to you and me can be incredibly difficult for AI technology to decipher. For example, if someone writes the word “park” in a tweet, is the word being used as a verb or a noun? Is the word “sex” being used as a male/female categorization or a description of a physical activity?
While there have been significant advances in brand safety classification and solutions over the years, the granular, semantic analysis of text is a subject most ad tech companies are just beginning to explore, says Anna Zapesochini, vice president of product management at DoubleVerify.
“The understanding of context is a sophisticated and ever-evolving space, requiring a combination of technology and human auditing and expertise,” she says.
DoubleVerify’s Semantic Science division is based out of Helsinki, led by a team of software development specialists with backgrounds in content classification. The new division is part of DoubleVerify’s integration of the technology it gained through the 2018 acquisition of the contextual intelligence and classification platform Leiki. Leiki’s software engine uses AI to provide high-definition analysis of text and contextual video data.
“Human expertise is guiding and informing that capability, while machine learning facilitates scale,” explains Zapesochini.
DoubleVerify plans to use the analysis data generated by Leiki’s AI engine to enable proactive contextual targeting of content that’s aligned with a brand’s equity or target audience profile. The goal is to protect the reputations of DoubleVerify’s brand clients throughout the media transaction, including pre-and post-bid.
In the coming years, DoubleVerify plans to extend the work of its Semantic Science team beyond text-based solutions, to cover brand safety and other media—including images, audio, and video. There are also plans to unlock new ad targeting and performance capabilities, including contextual targeting.
“We made this investment because we firmly believe that semantic analysis represents the future of brand safety and contextual targeting. Despite advancements in brand safety, the diverse nature of content on the internet requires constant vigilance and sophisticated tools to protect brand reputation,” Zapesochini says. “The combination of [DoubleVerify’s] technology and specialists within the Semantic Science division delivers on this need.”
Stephanie Miles is a senior editor at Street Fight.