"Rihe/Hello" starts from Minnan, Meta promotes voice translation

"Hi Mark, do you know that our team created the first translation system to support a spoken language?" (Hi Mark, do you know that our team created the first translation system to support a spoken language?)

Translating from Hokkien to English, you can rely on artificial intelligence to complete the translation only by speaking?

Meta President Mark Zuckerberg (Mark Zuckerberg) recently announced a video that their team is developing a set of Universal Speech Translator (Universal Speech Translator), which allows artificial intelligence to do oral translation and translate the language they speak into the other party's language.

The Hokkien language in this conversation is quite authentic. Researcher Chen Pengren grew up in Taiwan. His father is a southerner and is used to speaking Hokkien, so Chen Pengren also hopes that this technology can help more people like his father to communicate with others in a language they are familiar with.

However, the translation of Hokkien should take into account the differences between different regions. According to Meta statistics, there are about 28 million people who speak Hokkien in mainland Chinese, about 13.5 million people in Taiwan, and 1 million to 2 million people each in the Philippines, Singapore and Malaysia. 14,643 Hokkien translation tool will benefit Hokkien-speaking populations across Asia. (Taken from Meta)

On the other hand, Hokkien does not seem to have established a unified writing standard, and there are not many professionals engaged in Hokkien and English translation, which leads to very difficult challenges in translation work.

Therefore, the

Meta team borrowed the Chinese writing format as a reference, first translated Hokkien and English into Chinese, and then translated again. However, artificial intelligence is still unable to achieve real-time translation, and can only translate one complete sentence at a time.

Smart translation is paving the way for the meta-universe

Mark Zuckerberg's Meta team is working on developing translation tools, and in fact laying the groundwork for the establishment of the Meta Universe.

Considering that social interaction is the basic function of the meta-universe, people need to work and play in the virtual space. If the language barrier can be broken through intelligent translation, it will further pull people's distance in the meta-universe.

Meta's universal speech translation tool is an open source system, and it is hoped that real-time translation will be achieved in the future. It remains to be seen how long it will take to implement it.

A new direction for AI translation: unwritten languages

It is amazing to see artificial intelligence translate dialects, but its development may mean another possibility: preserving some unwritten languages.


Meta team pointed out that 40% of the more than 7,000 languages in the world are unwritten languages and do not have a standard writing system. As the population of speakers continues to decline, some languages are endangered. If they can be identified and translated by artificial intelligence, it will help preserve the diversity of human languages and cultures.

According to the "China's Endangered Languages" series published by the Commercial Press, among the more than 130 languages in China, 68 are spoken by less than 10,000 people, 48 are spoken by less than 5,000 people, and 25 are spoken by less than 1,000 people..

According to a UNESCO survey, there are actually about 90 endangered languages in the western part of the Austronesian language family (Western Astronesia), and some are extinct.

According to 2010 statistics, there are about 90 endangered languages in the western Austronesian language family. (Taken from UNESCO) The

Meta team revealed that they will develop the ability of artificial intelligence to translate other languages, and believe that artificial intelligence can solve communication difficulties for unwritten languages, which means that people may one day be able to communicate smoothly on the Internet in their own different languages. communicate smoothly.

Related articles