As per the news report on Monday, Facebook released software-based machine learning. The company said it was the first to be able to transform 100 of languages without relying on English. One of the researchers states in a blog post that this milestone is a conclusion of years of foundational work in machine translation.
Facebook Released a Software-Based Machine Learning for 100 Languages
A fan gave a review saying that the new model is more accurate and efficient than the previous other systems because it doesn’t rely on the English as the mediator translation step. Furthermore, a Fan wrote, “When translating, say, Chinese to French, most English-centric multilingual models, train on Chinese to English and English to
French, because English training data is the most widely available,”
Facebook said that on its news feed it is already handling more than 20 billion translations every day. It further hopes that the new system will deliver more and better results in the upcoming days.
“Breaking language barriers through machine translation is one of the most important ways to bring people together, provide authoritative information on Covid-19, and keep them safe from harmful content,” Fan said.
Facebook reveals that researchers used specific criteria for the selection of the language.
It includes different kind of families and those which are extensively used. However, they avoided different directions that are rare like Icelandic-Nepali or Sinhala Javanese.
However, they avoided statistically rare directions, including Icelandic-Nepali or Sinhala-Javanese. For the sake of different regions, the team organized languages into 14 major groups. These groups are further divided on the base of linguistic, geography and culture.
‘People living in countries with languages of the same family tend to communicate more often and would benefit from high-quality translations. For instance, one group would include languages spoken in India, like Bengali, Hindi, Marathi, Nepali, Tamil, and Urdu,’ Fan wrote.