The GIFGIF Website Translates Words Into GIFs, And Wants To Convert Shakespeare Next
Last week, MIT researchers created a water filtration system made from a pine branch... and now it's wowed us again. Two graduate students from the MIT Media Lab, Travis Rich and Kevin Hu, have created GIFGIF, a website that translates words into GIFs. How is GIFGIF doing that? Well, by quantifying the emotional content of a sentence or a phrase and matching it with a GIF that expresses that emotion.
Which sounds pretty difficult: How does a computer learn to understand emotional content in words? This is where humans come in. Upon arriving to the GIFGIF website, visitors are presented with two GIFs and with a question along the likes of, "Which better expresses anger?" The visitor selects the GIF she believes best expresses anger, and her choice gets added to GIFGIF's database.
With the help of Place Pulse, a data research team, Rich and Hu launched GIFGIF on March 3 — and since then, it's seen over 15,000 visitors a day, who vote on an average of 10 GIF pairs each site visit.
Still, creating an algorithm for computers to use based on this A/B voting system doesn't make translating text into GIFs any easier. As Rich told Quartz:
We’re already seeing that votes vary across different cultures and looking at which GIFs are the most volatile—which ones have votes change the most based on country—could help us understand how emotions are interpreted across the world.
You want to check it out now, don't you?
While Rich and Hu hope to one day create a computer that can translate a Shakespearean sonnet into GIFs — can you imagine? — that day could be a while off. Still, they've reported that ESL teachers are using GIFGIF to teach their students words for different emotions.
And here's our emotional response to what GIFGIF is trying to do.