Google has announced the expansion of its translation service by adding support for 110 languages using its PaLM 2 AI model. These languages, including Afar, Cantonese, Manx, Nko, Punjabi (Shahmukhi), Tamazight (Amazigh), and Tok Pisin, collectively represent over 614 million speakers, approximately 8% of the global population.
According to Google, the newly included languages vary widely in usage levels, with some having substantial speaker populations while others are maintained for cultural preservation despite limited active use.
The company emphasized its approach of prioritizing widely spoken variants of each language, taking into account regional varieties, dialects, and differing spelling conventions. For instance, Google’s models aim to produce text closely aligned with commonly used online variants, such as Southern Vlax Romani, while also incorporating elements from other dialects like Northern Vlax and Balkan Romani.
This expansion aligns with Google’s broader initiative announced in 2022 to support 1,000 languages through AI. Previously, in the same year, Google added support for 24 languages spoken by over 300 million people using its one-shot model.
Meanwhile, Apple recently announced at WWDC the addition of Hindi to its translation app, albeit supporting just over 20 languages compared to Google’s extensive language repertoire.
