Google Translate has expanded its support to include 110 new languages, including South Africa’s TshiVenda and siSwati.
In a statement, the internet search giant says this artificial intelligence (AI)-powered expansion is Google’s biggest yet, and is geared towards helping people better communicate and connect.
According to the company, part of Google’s 1 000 Languages Initiative, which uses AI models to support the 1 000 most spoken languages around the world, the latest expansion uses the PaLM 2 large language model. It follows the addition of 24 languages in 2022, which used Zero-Shot Machine Translation, it says.
This expansion represents the largest addition of African languages, with about a quarter of the newly-supported languages coming from the continent, says Google SA communications manager Siya Madikane.
Google notes that approximately 1.2 million South Africans speak TshiVenda, while siSwati, a Nguni language, is spoken by around 1.5 million people living in Eswatini and South Africa.
Their addition brings to nine the number of South African languages available on Google Translate, with Afrikaans, English, Sepedi, Sesotho, Xhosa, XiTsong and Zulu having been added in previous expansions.
The 110 new languages, which include Fon (primarily spoken in Benin), Kikongo (Republic of Congo, Gabon and Angola), Luo (Kenya and Tanzania), Ga (South Eastern Ghana) and Wolof (Senegal), represent more than 614 million speakers, making translation available to an estimated 8% of the global population, Google adds.
“Language is an essential part of how people communicate, understand and interact with the world around them, so we are excited about this latest expansion,” says Madikane.
Google points out that among the other languages included in the latest rollout, are:
- Afar, a tonal language spoken in Djibouti, Eritrea and Ethiopia. Of all the languages in this launch, Afar had the most volunteer community contributions.
- Cantonese is one of the most requested languages for Google Translate. Because Cantonese often overlaps with Mandarin in writing, it’s tricky to find data and train models.
- Manx is the Celtic language of the Isle of Man. It almost went extinct with the death of its last native speaker in 1974. But thanks to an island-wide revival movement, there are now thousands of speakers.
- NKo is a standardised form of the West African Manding languages that unifies many dialects into a common language. Its unique alphabet was invented in 1949, and it has an active research community that develops resources and technology for it today.
- Punjabi (Shahmukhi) is the variety of Punjabi written in Perso-Arabic script (Shahmukhi), and is the most spoken language in Pakistan.
- Tamazight (Amazigh) is a Berber language spoken across North Africa. Although there are many dialects, the written form is generally mutually understandable. It’s written in Latin script and Tifinagh script, both of which Google Translate supports.
- Tok Pisin is an English-based creole and the lingua franca of Papua New Guinea.
Madikane says a lot of consideration goes into new language additions for Google Translate, ranging from which languages to include, to the use of specific spellings.
“Many languages do not have a single, standard form, so learning the specific dialect that is spoken the most in an area is more feasible. Our approach has been to prioritise the most commonly used varieties of each language,” he says.
PaLM 2 played a crucial role in assisting the team to translate more efficiently and learn languages that are closely related to each other, says the internet search firm.
Google explains that as advancements in technology continue, it will partner with expert linguists and native speakers to support more language varieties and spelling conventions over time.
Share