Thursday, July 25, 2024 Banner
HomeSoftwareAIGoogle Translate's Largest Expansion Ever: 110 New Languages Added Using AI.

Google Translate’s Largest Expansion Ever: 110 New Languages Added Using AI.

Strong artificial intelligence (AI) algorithms were employed, as Google disclosed on Thursday, to enhance its translation platform and offer 110 new languages for users.

The biggest extension of Google Translate to date has been made possible by the IT giant’s implementation of the PaLM 2 large language model (LLM).

Cantonese and Q’eqchi are among the new languages that represent about 614 million speakers, meaning that almost 8% of the global population can now have their translations available, according to a press statement from Google Translate’s senior software engineer, Isaac Caswell.

More than 100 million people speak some of the major world languages, according to Caswell. Some languages are spoken by tiny Native American populations, while others have very few native speakers, despite ongoing revitalization initiatives. The largest growth of African languages to date, comprising Fon, Kikongo, Luo, Ga, Swati, Venda, and Wolof, accounts for almost 25% of the new languages.

One of the notable features that Google Translate’s statement emphasized is Cantonese. Although adding Cantonese was challenging due to its frequent writing overlap with Mandarin, which makes it challenging to collect data and train models, Caswell remarked that it has long been one of the most requested languages for the program.

Shahmukhi, the most widely spoken language in Pakistan and a Punjabi variety, and Afar, the tonal language spoken in Djibouti, Eritrea, and Ethiopia, were also added. According to the press release, Afar made the largest contributions to the community as volunteers.

Google said it plans to progressively add more languages to Translate as part of its previously announced 1,000 Languages Initiative, which commits the company to creating AI models supporting the 1,000 most widely spoken languages in the world. The corporation claims that as AI technologies like PaLM 2 advance, that process will quicken even more.

According to Caswell, PaLM 2 was a crucial component in enabling Translate to acquire closely related languages more quickly. These languages included French creoles like Seychellois Creole and Mauritian Creole, as well as languages similar to Hindi like Awadhi and Mardwadi.

We’ll support more language variations and spelling standards in the future as technology develops and as we work with linguists and native speakers, he continued.


Most Popular

Recent Comments