Over the last few years, it’s really felt as if artificial intelligence has come on leaps and bounds. Amazon’s Alexa can provide you with real-time traffic information and Apple’s Siri can stream your favourite music – now you can even have a virtual interpreter in your pocket. Let’s have a look at where machine translation started and if AI is really going to play a major role in the translation industry.
Machine translation’s influences can be traced right back to even the Second World War and the revolutionary code breaking. It didn’t take long for proposals to be submitted and in 1954, the Georgetown-IBM experiment made headlines for automatically translating more than sixty Russian sentences into English. Although it was impressive for its time, it was just as limited as the system was tailored for organic chemistry and included just six grammatical rules and 250 lexical items (an element such as a word or a phrase that gives meaning).
Despite the coverage of the Georgetown-IBM experiment, research into machine translation in the US was halted in the 1960s. The Automatic Language Processing Advisory Committee was assembled by the US government and the ALPAC report followed, concluding that MT wasn’t likely to compete with a human translator and was essentially more expensive and less accurate. As globalisation in the following decades sped up, the demand for translation services increased along with the amount of systems used for translation. The use of CAT tools became much more common in the industry and Babel Fish and Google Translate became household names.
Google’s virtual assistant – inventively called Google Assistant – has been around since its release in 2016 and its interpreter mode was showcased at the Consumer Electronics Show in Nevada earlier this year. Although the Assistant’s Interpreter feature may sound futuristic, it has already been rolled out for 29 languages and hotels tested it in its pilot phase. A video shown to press in Las Vegas shows the Google Assistant in action at one of these hotels. We see it break down the language barrier between German and English speakers without any headphones in sight and, perhaps most usefully, it eradicates the need to contact suppliers before they are actually needed since the device is already accessible, and at a one off price once you’ve bought a Google device.
Limitations of MT
While this is all very promising, there are still a few limitations that AI presents when it comes to language and translation. So far we have only seen the Google Assistant interpret a fairly basic conversation between a guest and a receptionist. While this got the job done, reports of people using the function for more spontaneous conversations suggest that it unsurprisingly causes their communication to be stilted – larger volumes of more technical language may pose more of a problem for the Interpreter. Research and understanding context are important parts of translation and something our translators at AST are excellent at having lived abroad or travelled extensively, not to mention having translated for a range of different clients with different backgrounds and requirements – virtual assistants don’t have this advantage so they can’t alter their language to better suit target audiences, consider cultural sensitivities or master a number of specialisms.
The pioneers of the Georgetown-IBM experiment might not have envisaged that in 2019 we’d be given the opportunity to break down language barriers with just our mobile phones – in fact, they probably didn’t expect mobile phones to be so advanced at all. It’s difficult to predict how technology is going to continue to influence the translation industry, but we’ve seen how far it has come already and we don’t doubt that it has much further to go.