11 Apr Rebuilding the Tower of Babel: Why we’re 15 years away from understanding every language
From floods to fire and brimstone, the Bible’s Old Testament is full of inventive ways that God admonished the early humans for their sins… but what happened at the Tower of Babel was perhaps the most enduring. According to Genesis, once back on dry land from Noah’s Ark, mankind – united in a single language – built a tower to reach heaven so they could be likened to God. To punish them for their arrogance, God destroyed the single language and flung the people to every corner of the earth in confusion, marking the last time that all of humanity was united in a single goal.
Thousands of years later, Professor Roberto Navigli from Sapienza University of Rome is aiming to help rebuild that tower – not with bricks, but with computing power. His latest project, aptly named BabelNet, is attempting to unite more than 280 languages with artificial intelligence. Rather than a translator simply matching words (often badly), the aim of BabelNet is to actually understand languages when it compares them.
“The system can potentially read the whole web. It can tell me about a book written in a language that I can’t read and then I’ll be able to ask it questions to get the gist of it,” he explains.
It’s a future that is not too far away – between 10 to 15 years, according to Roberto. His programme works with dictionaries, which he describes as like islands; full of life and beauty but ultimately disconnected from one another. But if you can begin to connect these islands, joining up the thousands upon thousands of print and digital dictionaries in existence, then you could create a system that understands every element of human speech.
What does the existence of such a system mean for society? First, it means you get to have the last laugh over every language teacher who told you that without French/Spanish/German etc. you’d be unemployable. But more broadly, it brings all of society closer together by facilitating conversations between anyone on the planet.
“My idea is to create representations for sentences that are independent of the language from which they were obtained. I could show you symbols, pictures and the various concepts that are conveyed in a sentence, interconnected in a graphical representation that you could understand even if you don’t speak that language because that representation is not language dependent.
“Maybe you can then transform it into a sentence in your language. That is my dream,” he says.
It’s possibly the biggest challenge the field of linguistics has ever faced. By its nature, language sidesteps mathematical theories and rules. Sarcasm, ellipses, slang, double entendre and dishonesty are just a few charming human traits that can confound a computer. Those are just within one language – but humanity’s languages are also wildly different from one another. “There are no spaces in Chinese. Even understanding what a word is, where it starts and where it ends is a challenge. There’s a lot of implicit disambiguation that is done automatically in the human brain, but for a computer, this is very difficult.”
Roberto and his team are getting closer. We may one day be the last generation to remember a time before anyone could speak to anyone else. And for a species that thrives on communication, this is certainly a noble pursuit. The foundations for the new Tower of Babel have already been laid, only this time the ‘heavens’ that we will reach will be the next giant leap for mankind.