Millions of languages exist, but translating them all seemed like a far-off dream—until now. Google Translate bridges communication gaps across hundreds of languages, but a vast number of lesser-known ones remain untranslated.
That changes today as Google Translate is leveraging the power of generative AI to tackle these previously untranslatable languages. The company has announced support for a whopping 110 new languages, significantly expanding the reach of communication and understanding.
The list of new languages includes some familiar names like Cantonese, a top user request for Google Translate. Primarily spoken in parts of South China, Cantonese shares the same writing system as Mandarin. In the past, this overlap created a hurdle for traditional machine-learning models. They struggled to distinguish the nuances of each language based solely on the written characters. Generative AI finally helped bridge that gap, allowing Google Translate to accurately capture the unique aspects of Cantonese.
Imagine trying to translate between Southern and American English. They seem similar but have subtle differences. Thanks to PaLM 2, its powerful generative AI model, Google Translate can handle closely related languages.
Languages like Awadhi and Marwadi, both related to Hindi, were also challenging. PaLM 2 tackled this issue by analyzing vast amounts of text data to identify the unique nuances of these closely related languages. This opens the door for more accurate translations for millions of people who speak these languages.
With sights set on at least 1,000 languages, Google Translate aims to bridge the communication gap for billions. While an ambitious goal – the estimated 7,000 languages spoken worldwide present a significant challenge. AI advancements like generative models are chipping away at this challenge, allowing Google Translate to distinguish between regional dialects and variations within a single language.