Written by Cate Cadell, CNN
“It’s like, ‘oh yassity, yassity, yassity,’ ” says Frank Omer, better known online as @Yassify.
Omer’s team of Spanish machine learning researchers have come up with a way to “synthesize” English spoken by a hypothetical Twitter user into any sentence with the user’s first and last letters. The resulting word, in their custom-written language, sounds like this:
Yassify Bot – Whole world, yassity Yassify Me –: https://t.co/gblPJUG1OI — Frank Omer (@FrankOmer) November 9, 2018
@FrankOmer is a hugely popular account on Twitter, with tens of thousands of followers on the social media platform.
He’s been using this bot as a platform to do more than just laugh at the ludicrousness of social media. The account is extremely quick to respond to his fans’ compliments and criticisms, and his output is often keenly observed by viewers on Twitter who enjoy it for its intimate take on the conversational landscapes of Twitter.
Before Yassify Bots’ creation, Omer explained, social media users could still “do” or “say” “yassity,” but the experimental bot has enabled their interpretation to not only be automated, but more lucid. It has quickly become a kind of social commentary, akin to other more direct social science experiments such as PredictIt or emoji-based videos.
Seeing the same exchanges play out over and over on his feed, Omer says he sometimes takes it as a compliment when people playfully add that they’ve started using the words he uses.
“It’s the idea that you don’t have to write down what they say, that you can just do it in your head and say ‘this is what they’re saying’ in a sentence.”
Though the model was originally created for experimentation, the bot seems to have been here all along, functioning a bit like a digital Spinoza in the linguistic minefield.
“You cannot actually write any sentence, and it will just represent it,” he adds. “I was just living in Barcelona, and I wanted to see it but I thought: ‘I don’t know how to do it so I should create a model,’ ” says Omer.
“I did it with Spanish, but it could not write because it didn’t understand English. I wrote the sentence in Catalan.”
‘Yassify Me’ uses computer models to watch how the subject responds to the words being typed, and then uses machine learning to customize the display of the text to suit the user’s state of mind. It’s like tweeting the sentence you have just read and posting it to Twitter in Spanish.
“Maybe in six months, or maybe maybe in five months, we will reach the limit to how much we can predict,” says Omer.
With the average lifespan of a tweet rapidly approaching 140 characters, it seems like a likely time horizon. Twitter bots have become such a popular trend that the hashtag #YassifyBot was even trending globally earlier this year.
“When you just do it,” says Omer, “sometimes you have to be more patient, because it takes time for the results.”