Google Translate adds 110 new languages in largest expansion ever: Awadhi and Marwadi among others
American tech giant Google has significantly expanded its Google Translate service, adding support for 110 new languages. This major upgrade, announced on Thursday, marks the largest expansion in the history of the translation tool, which now includes a total of 243 languages.
The latest additions were made possible through the use of Google’s in-house PaLM 2 large language model (LLM), which utilized artificial intelligence (AI) to learn new languages, particularly those that are related or regional dialects.
Among the newly supported languages are Indian regional dialects Awadhi and Marwadi. This expansion is part of Google’s ambitious 1,000-language initiative, announced in November 2022, which aims to create AI models capable of supporting the 1,000 most spoken languages globally. The inclusion of these 110 new languages represents a significant milestone in this ongoing project.
Google prioritized the inclusion of the most commonly used varieties of each language. For instance, when considering the Romani language, which has numerous dialects across Europe, the AI model selected the Southern Vlax Romani dialect, as it is widely used online. The availability of textual data for training the AI model also influenced the selection of new languages.
Some of the notable additions include Cantonese, one of the most requested languages for Google Translate, and Afar, a tonal language spoken in Djibouti, Eritrea, and Ethiopia. A significant portion of the new languages, approximately one-fourth, are from Africa, reflecting Google’s commitment to linguistic diversity and inclusion.
In a blog post, Google emphasized the role of volunteer community contributions in identifying and adding new languages, particularly in regions where digital resources are limited. This expansion underscores Google’s ongoing efforts to bridge language barriers and make information accessible to a broader audience.