Providers are now beginning to overcome the practical hurdles of this method, chiefly the huge computational resources and time it requires.
Google launched its ‘Google Neural Machine Translation (GNMT) system’ last September which utilises “state-of-the-art training techniques to achieve the largest improvements to date for machine translation quality”, the company wrote in an announcement.
Using bilingual-human-rated, side-by-side comparisons of SMT and GNMT translations of sampled sentences from Wikipedia and news websites, GNMT has reduced translation errors by more than 55 per cent to 85 per cent in several major languages.
In November, Microsoft too adopted NMT, providing “major advances in translation quality” over SMT technology, the company said.
Amazon Web Services is also expected to make machine translation services available this November – according to a CNBC report – building on its AI push, and likely using NMT. Facebook’s Mark Zuckerburg in May outlined his company’s machine translation research using convolutional neural networks, which work in a slightly different way. “With a new neural network, our AI research team was able to translate more accurately between languages, while also being nine times faster than current methods,” he wrote on his Facebook page.
There are still significant limitations. Microsoft said use of neural networks was “only a first step towards future improvements”. Google agreed, saying of their advances: “Machine translation is by no means solved”.
With constant refinement, these techniques are however “getting pretty close to human level performance,” said Norvig, at an AI event at UNSW in Sydney last month.
But even if machine translation can beat human translators, the results will still be far from perfect.
“One thing we’re finding is that human level performance isn’t always that great. Human translators make mistakes too, because they’ve got a deadline, they’re going fast, they didn’t always understand the source material and so they said the wrong thing. So maybe it’s not our goal only to reach human performance, maybe we should try to correct all of the mistakes they make as well,” Norvig said.
Perfection, according to Norvig, will come from giving the machines a new skill: real world wisdom. “There’s some things where you really need to understand the world in order to get the right answer,” he says.
Bowling ball barriers
Norvig makes his point with this example: “If I say ‘I dropped the glass on the table, it broke’, then you understand that it means the glass. And if I say ‘I dropped the bowling ball on the table, it broke’, then you understand it’s probably the table that broke. And that doesn’t have to do with linguistics – that has to do with physics and the real world. To get those kinds of things right, we’re going to have to teach our machine translation system what the world’s like.”
Sign up for CIO Asia eNewsletters.