The recent inclusion of the Tshivenda language in Google Translate has been warmly welcomed by the people of the Vhembe District. Google Translate has made significant strides in overcoming language barriers through its latest expansion, driven by artificial intelligence.

With the addition of 24 languages in 2022, using Zero-Shot Machine Translation, Google Translate has now included Tshivenda and siSwati, bringing the total number of South African languages offered to nine. Previous expansions included Afrikaans, Sesotho, Sepedi, isiXhosa, XiTsonga, and isiZulu.

Tshivenda, spoken by approximately 1.3 million people primarily in the Limpopo Province, with the majority in the Vhembe district, has been integrated into the Google Translate platform alongside 110 other languages. This addition underscores the importance of preserving linguistic diversity and empowering speakers of minority languages.

Dr. Naledzani Rasila, former chairperson of the Provincial Language Committee and current director of monitoring and evaluation at Limpopo Education, welcomed this development. He said this was an important step for communities where Tshivenda may not be widely understood beyond their immediate region.

“This will enable the younger, more technology-oriented generation to connect more easily with their language. It will aid in the preservation of language and culture, although we need to exercise caution regarding cultural aspects like dance, song, dietary practices, and lifestyle, while avoiding matters related to rituals and initiation schools,” he said.

Rasila also raised concerns about potential threats to the language from incorrect spellings, expressions, and the misuse of idioms and proverbs.

Siya Madikane, Google SA’s communications manager, explained the significance of this AI-powered expansion as Google’s most extensive effort yet to enhance global communication and connectivity. “As part of Google’s 1,000 Languages Initiative, utilising AI models to support the world’s most spoken languages, this expansion utilizes the PaLM 2 large language model. PaLM 2 significantly enhances translation efficiency and language learning, especially for closely related languages,” Madikane explained.

Khathutshelo Raliphaswa, who translates articles from English to Tshivenda for Limpopo Mirror, was optimistic about the possibilities that the new service could unlock. “Since I started translating, I have faced one significant challenge, which was constructing sentences properly using Tshivenda. I hope this new tool will assist me with translation,” she said.