Many may not realize that there are 7000 languages still in use today but 40% of them are purely oral without a written system. Therefore these languages need to be used to prevent them from becoming extinct. Among the languages that do not have an official written system is Hokkien. All transliteration systems in use today require a written system to function. Therefore if you want to translate the Hokkien language there is a problem.
Peng-Jen Chen a researcher at Meta AI wants to solve this problem because he wants to help his father who is more fluent in Hokkien. Throughout his life in Taiwan, Chen realized that his father had difficulty interacting using Mandarin. Therefore an artificial intelligence that can translate Hokkien was developed by him.
The system is first trained using English and Mandarin as an intermediate to Hokkien. Mandarin is used because it has a lot in common with Hokkien. The translation process is carried out by Hokkien language users throughout the process to ensure accuracy. Then artificial intelligence is trained using this data set.
Besides Hokkien, Meta believes the same artificial intelligence can be trained using languages that have no other written system. This is one of Meta's efforts to produce a Universal Speech Translator - a universal translator that can translate all the world's languages in real time.