Meta, the parent company of Facebook, announced Wednesday that its AI systems in this field are now capable of fully translating two hundred languages interchangeably, whatever the structure, after it was limited to 100 languages. Meta founder Mark Zuckerberg said in a blog post that “many” of the languages involved in this new advancement “were not available for machine translation.”
This announcement is an example of the machine translation rush among major Internet groups trying to deliver their services and products to the entire population of the world. In May, Google researchers published a scientific paper titled “Building Machine Translation Systems for the Next Thousand Languages,” reflecting the big ambitions in this field.
The artificial intelligence systems developed by Google, Microsoft and Meta are now capable of translating languages for which little parallel data is available, which means they are rarely translated into other languages.
It allows for the translation of texts into languages with limited reach on the planet, such as Quechua (which is largely confined to Hebero) and the language of the Fulani people (in West Africa), although no human has ever been involved in this task. Francois Yvonne, a researcher at the National Center for Scientific Research in France who specializes in linguistics, says that machine translation is “especially important to Facebook, which has to monitor the hate messages” that are spreading around the world and in all. languages. He explains that machine translation can allow, in particular, English-speaking moderators to interfere with content posted in other languages.
But the reliability of these tools remains to be determined. In this context, Meta points out that its new system is able to provide “44% better performance than the previous model, which was capable of translating 100 languages. In some African and Indian languages, this improvement exceeds 70% compared to” to the recent use of translation systems, According to the world’s leading social network.
But François Yvonne believes that the quality of the automatic translations provided by Google or Facebook engines will undoubtedly remain unequal depending on the language.
Speaks 200 languages
“Highly translated languages, such as European languages, are likely to retain the lead,” he says.
The same opinion is shared by Vincent Godard, CEO of the main French machine translation company, Systran, who for his part operates in 56 languages. Godard points out that the technology used by this group is initially the same as Meta and Google, but has been enhanced by the work of real linguists to avoid mistakes. “When you translate a fighter jet assembly manual, you can’t allow a single mistake,” he explains, while mistakes can be accepted when it comes to translating an opinion on a restaurant.
So, are we closer to having automatic voice translation, to be able to speak directly to anyone on this planet, for example in the future world of Metaverse? “We haven’t gotten to that point yet, but we’re working on it,” says Antoine Bord, general manager of Meta Network’s VAIR Research Laboratory for Artificial Intelligence. “We have another project on automatic speech translation, which is currently working in far fewer languages,” he says. But the advantage will be “linking the two projects, so that one day two hundred languages can be spoken while maintaining the tone of speech, emotion and dialects …”, according to Burd.
Read More About: Technology News
