Eight percent of people who viewed the video of former child star Corey Feldman’s recent surreal, Michael Jackson-inspired “Today” (NBC) show performance on YouTube found it to be “super-awkward.”
That’s according to data from Canvs, a language analytics platform that measures emotions on social media. Its automated system not only factors in the likes and emojis (such as “love,” “sadness” and “anger” symbols), it also interprets written comments and what company co-founder Professor Sam Hui calls “otherwise incomprehensible millennial slang,” and maps the those responses to a range of 56 emotions.
“About 65% of the emotions we express in social we express are not in the dictionary,” says Canvs co-founder Jared Feldman. “They’re made-up words and misspellings of made-up words, they’re strange phrases that didn’t exist a couple of months ago.”
Canvs previously offered insights on Twitter posts. Today, it officially adds YouTube and Facebook to the mix.
“If you’re an influencer on YouTube, and you’re constantly putting out new content, you can see all your clicks and your views, but you don’t know why it’s happening or who you should be engaging with it,” says Feldman.
Feldman met Hui when he was a student at NYU and Hui was an assistant professor of marketing in its Stern School of Business, and they went on to found the company Mashwork, which first rolled out Canvs in 2014.
Prior to creating Canvs, “we used third party social listening tools to create insights for brands, mainly in the media and entertainment space,” says Feldman. “HBO has shows that premiere Sunday night, and Monday morning they’d say, ‘How did we do last night?’ A totally reasonable question. When it came to figuring out how many tweets they got, how many comments, how many shares, etc., etc., it’s super easy. When it came to figuring out what percentage of people were upset that Joffrey died on ‘Game of Thrones,’ it would take until Wednesday or Thursday, because we’d either have to manually code tweets or use supervised systems that required hours of training, and even then if I wanted to know what was the funniest moment on TV last night, it was literally impossible to figure out.”
Today, Canvs’ clients include, brands, agencies, studios, multi-channel networks (including StyleHaul) and two thirds of the U.S. TV networks.
For live nature content producer is Explore.org, it analyzed more than 280K Facebook comments on its content from June through mid-September and found that 42% expressed love, 28% beauty, and 16% joy.
Canvs also did an analysis of the social chatter on YouTube and Twitter for the upcoming major network shows with the most activity and found the top five shows ranked by percentage of comments that expressed “love” and “excitement”
Feldman says the emotional measurements are deeply important, “because, as humans, 90% of our decisions are driven by emotion, and 95% of all emotional-based research is still done face-to-face, in telephone interviews and focus groups,” which are a $40 billion industry.
Canvs receives its Twitter data from Nielsen, which measures program-related Twitter activity for linear episode airings on a 24/7 basis.
To help keep its system up-to-date on the latest slang, it developed a machine learning application it calls the Judgment Box. A network of people from around the country is periodically called upon to log into its dashboard and answer a series of yes or no questions.
“It’s like Tinder, but for emotions,” says Feldman of the Judgment Box. “You’re basically saying ‘I agree with this’ or ‘I don’t agree with this.’” When it comes to understanding the latest slang, “it turns out that millennial females are really good at deciphering ‘Real Housewives of New York’ conversations, whereas a different type of human is better at understanding sports and all the nomenclature and nuances there.”