Microsoft has successfully used artificial intelligence (AI) to match human performance in translating news from Chinese to English, according to a recent blog post by Allison Linn, senior writer and editor at Microsoft.
Researchers in the company’s Asia and US labs said that their system achieved human parity on a commonly used test set of news stories, called newstest2017, which was developed by a group of industry and academic partners.
“Hitting human parity in a machine translation task is a dream that all of us have had, we just didn’t realise we’d be able to hit it so soon,” said Xuedong Huang, a technical fellow in charge of Microsoft’s speech, natural language and machine translation efforts.
Academic and industry researchers have recently achieved substantial breakthroughs by using a method of training AI systems called deep neural networks. This has allowed them to create more fluent, natural-sounding translations that take into account an even broader context than the previous approach, known as statistical machine translation.
To reach human parity on this dataset, three research teams in Microsoft’s Beijing and Redmond, Washington, research labs worked together to add a number of other training methods that would make the system more fluent and accurate.
One method they used was dual learning which involved sending a sentence through the system to be translated from Chinese to English, the research team also translated it back from English to Chinese.
Another method, called deliberation networks, was used to teach the system to repeat the process of translating the same sentence over and over, gradually refining and improving the response.