Found in translation uses machine learning to show language’s similari

For those who’ve ever traveled to a overseas nation and also you didn’t converse the language, you in all probability had your share of conversations that acquired misplaced in translation. Language limitations usually stop shared understanding. A brand new set up in an exhibit at 21_21 Design Sight in Tokyo flips the thought on its head.

The interactive set up “Found in Translation,” produced by the design studio TheGreenEyl with Google Artistic Lab and Waseda College affiliate professor Dominiq Chen, uses knowledge visualization and different audio-visual options to show what totally different languages have in widespread—and to have a good time what makes every language distinctive.

The set up encompasses a single podium surrounded by giant screens. Upon coming into, the customer solutions a immediate. Their reply is then pushed by Google Translate, and audio of the sentence begins to feed by audio system in 24 totally different languages. Then, the sentence seems as a knowledge visualization on screens all through the room. Lastly, the sentence splashes on screens throughout the room in giant, all-caps kind. It’s a poetic method to show that we aren’t so totally different in spite of everything.


The information visualization element, which reveals the sentence you simply spoke as colourful, squiggly strains, uses machine learning to convey the semantic similarities between languages. The strains seem closest collectively the place the person phrases are most related, and farthest aside the place they’re probably the most dissimilar. (A full clarification of how this works requires greater than you in all probability want to learn about machine learning, however you may learn the details here). They’re additionally color-coded by language household so the sentence you spoke, say, in English, will seem as one colour, and the translated sentence, say in Japanese, will seem as one other. Visible similarity relies on how a lot the 2 languages have in widespread, however the line graphs additionally provide an enchanting inside take a look at language construction total—and the way differing languages might need extra in widespread than you’d suppose at first look.

[Photo: Taiyo Watanabe/courtesy TheGreenEyl]

After the info visualizations fade from the display, the customer’s response seems written out in a rainbow of various colours and languages throughout the screens. Whereas the info visualizations seize the empirical similarities of language, the sort remedies add a human contact—even in their clear, sans serif type, in accordance to Richard The, artistic director and accomplice of Studio The GreenEyl and the lead on this undertaking. “It’s very a lot a typographic undertaking the place you see totally different writing programs from around the globe, and on the similar time it really seems like a really human second the place you out of the blue hear instantly how all people could be talking this phrase or sentence from around the globe,” The says. “It’s an enchanting and heat second.”

The set up was initially developed to run concurrently the Tokyo Olympics in 2020—however when the COVID-19 pandemic put the world as we all know it on pause, Found in Translation paused too. After The and his staff spent months remotely establishing the set up as a result of they couldn’t journey to Japan, the interactive set up is now on view in Tokyo through June 13.