Do you use machine translation to learn a foreign language? Large data cannot replace artificial

Machine translation error joke a lot of jokes, but the application of large data makes this technology increasingly perfect. Even proponents believe that machine translation will eventually replace the human race to overcome language barrier communication. But can technology really fathom the richness and subtlety of human language?

The primary machine translation is more like an electronic dictionary, is good at translating individual words and simple sentences, but it is helpless for polysemy. That is why, despite years of development, Pure machine translation is still riddled with errors. The advent of the big data age could completely rewrite the situation.

The more text data is entered, the higher the accuracy rate of machine translation. In different languages, creating a machine translation pattern usually requires 30 million words or 100多万个 sentences. Thanks to the presence of institutions such as the United Nations and the European Union, its translators have accumulated rich parallel corpus (also known as translation memory) for many years to save a lot of trouble for machine translation researchers.

It also explains why Google has performed well in machine translation, with its huge data and information. Google's machine translation proprietary system is a model of the industry, to achieve 63 of the common language of interactive translation.

However, human language is so complex and subtle that it is far from being a parallel corpus. If there is a metaphor or any poetic language, it is very difficult, experts say, if the use of a translation system has never seen a pun, the computer will only literally literal.

If we consider the provenance of various languages and cultural differences, some words can not be translated, that is, the so-called untranslatability phenomenon, you can not expect machine translation to replace human translation.