Google Translate recently turned ten. Their statistical model (based on what people actually speak and write) has won over a great deal of the market from traditional rules-based linguistic models (how people should actually speak and write), though perhaps new hybrid models may win the day in the long run.

The past decade has seen a revolution in the capability for machine translation (MT), with such heavy hitters in the tech world as Google and Microsoft one-upping each other in terms of providing (generally) free and open translation services to the public (while vendors like Lilt, Trados and MemoQ focus on the professional market). Whether embedding MT into social media and apps, such as Microsoft’s alliance with Twitter or its integration with Skype, to immediately translating street signs from pictures using Google’s Word Lens technology. Or enabling developers with developer APIs such as for Microsoft Translate.

Such “overnight success” for the MT industry wasn’t “overnight.” It’s been a series of step-by-step advancement over the decade. Or, actually, multiple decades, if you go back to the advent in 1997 of AltaVista’s (later Yahoo’s) BabelFish.

Google was practically a decade late to the game, but caught up quickly. Like Altavista before it, Google initially began by using SYSTRAN’s engine under the hood. But then, applying a great deal of internal R&D resources, it turned machine translation from a peripheral neato technology featurette to a central — and internally-developed — offering. Along the way they also acquired technology, like the Word Lens feature from Quest Visual, and cleverly integrated open source where appropriate, such as for their wordnets.

Google Translate is still a work-in-progress. And, yes, it still needs a lot of work. We’ve already highlighted how MT can munge timeless prose or lyrics, such as in Google Translate Sings, or how Hamilton creator Lin-Manuel Miranda did not mean to say the future will be “bouncing.” While MT has come a long way, it is still not at the same par as a naturally-fluent individual, never mind a professional translator, especially when dealing with nuance, context or even basic grammar.

If anything, though, Google Translate and its other publicly-available service peers have primed the global audience for translation services. People have come to expect web pages to be available in their own language, and do get slightly irked when the information from free and open MT tools comes off sounding odd. Marketers now have distinct choices: to leave their brands and their message to the vagaries of the online translation tools, or to take matters into their hands proactively to make sure their message comes across crystal clear.

What are your thoughts and experiences on Machine Translation vs. human translation? We’d love to know! Email us at [email protected].