Microsoft's New Machine Translation System Accurately Translates Chinese to English

A team of Microsoft researchers has reportedly developed the world’s first machine translation system that can accurately translate Chinese news articles.

On Wednesday, Microsoft researchers announced that they have created a machine translation system capable of translating Chinese news articles to English with the same level of accuracy as humans.

In a blog post published yesterday, the researchers from the company’s U.S. and Asia labs claimed that their system has already achieved human parity when tried on a commonly used test set of news stories known as newstest2017. The said test was released last fall at the WMT17 research conference by a team of academic and industry partners.

“Hitting human parity in a machine translation task is a dream that all of us have had,” Huang Xuedong, in charge of Microsoft’s speech, natural language and machine translation efforts, was quoted as saying. “The pursuit of removing language barriers to help people communicate better is fantastic.”

According to Xuedong, their machine translation system is considered a “major milestone in one of the most challenging natural language processing tasks.” He further stated that it’s a gratifying feat because of its potential to help people better understand each other.

Microsoft’s Machine Translation System

Compared to speech recognition, which has seen several improvements in the past number of years, teaching a machine to understand language has been a more challenging task for researchers.

To date, advances in artificial intelligence and speech recognition have paved the way for voice assistants to reach our phones and our homes. However, getting a machine translation of a web page or news article still yields messy and unclear results up until now.

In most cases, for a news article to be accurately translated, a human translator must do it. However, even human translators have slight differences in translating foreign words.

“Machine translation is much more complex than a pure pattern recognition task,” Ming Zhou, assistant managing director of Microsoft Research Asia and head of a natural language processing group that worked on the project, said. “People can use different words to express the exact same thing, but you cannot necessarily say which one is better.”

For the researchers to achieve their parity milestone, they reportedly added a few other training techniques to make their machine translation system more fluent and accurate. The methods were said to be based on how humans improve and develop their work by repeatedly doing it over and over until they get it right.

According to the researchers, the new methods they used on the system include dual learning for fact-checking translations, deliberate networks for refining translations, and joint training for English-to-Chinese and Chinese-to-English translation systems boost. There is also an agreement regularization to enable the generation of translations through reading, either from left-to-right or vice versa.

Right now, Microsoft has not yet made any announcement on what they are planning to do with the new machine translation system as it is still under study.

“We’re working to bring this to production as soon as possible, but we have nothing to announce at this time,” a spokesperson for Microsoft was quoted as saying.

“In the future, these systems could be applied to Microsoft’s commercially available translation tools such as Microsoft Translator, which is available as an app, API, and also the translation engine for many Microsoft products including Office, Bing, and others.”