At the same time, many languages remain unwritten or poorly written, especially those of small indigenous peoples who retain their traditional way of life. Sometimes an unwritten language can be called a dialect with a controversial status.
Almost half of the 7,000 languages known in the world and 4 out of 10 languages are spoken directly without being written down. These unwritten languages cannot be translated, especially in translation systems of digital algorithms. Working on the “Universal Speech Translator”, Meta announced in July 2022 that it is working on 200 languages that are difficult to translate, such as Asturian, Luganda and Urdu, within the “No Language Left Behind (NLLB)” project. The meta-researchers studied the unwritten Hokkien language, one of Taiwan's official languages, spoken by minorities in Asia.
Trying to create algorithms for translating unwritten languages, Meta AI normally processes machine translations using both written and spoken examples of the language.
Why is it necessary to focus on unwritten languages?
Because by many algorithms, languages such as English, Spanish can be translated with high accuracy. However, according to the data shared by Meta, the languages used by 20 percent of the world's population are not supported by these systems.
