African languages left behind by AI, UBC researcher works to close gap

A University of British Columbia (UBC) researcher is working to build an artificial intelligence (AI) system to ensure African languages aren’t left behind.

Systems like ChatGPT, which is an AI tool that generates text and interacts with a user in a conversational manner, are only offered in English.

Ife Adebara, a UBC Linguistics PhD student and the lead researcher on the project, says many of the current AI systems available rely heavily on language data sets that are widely available, like English.

“This volume of data does not exist for many African languages, so what I do is build data sources and collect sources of data to build technologies for African languages … to ensure [these] languages [do not get] left behind,” she said.

“I am able to see the importance of having access to technology in languages [people] understand and speak. Most Africans do not have an adequate understanding of English and French and other languages that already have enough AI technology.”

AfroLID

Adebara says she and her team have built a Natural Language Processing (NLP) system — which teaches computers to process human language — for African languages known as AfroLID. She says AfroLID is able to take a random excerpt and identify the specific language it is from.

“For example, Alexa uses NLP to have conversations with people. Google Translate uses NLP to understand a source language and translate it into a target language,” researchers said in a release.

Adebara says her system is the first step in building a chat model such as ChatGPT for these specific languages.

She tells CityNews her motivation behind the project comes from the increasing importance these tools play in every day life.

AfroLID currently supports 517 African languages with about 95 per cent accuracy which, Adebara says, supports the largest number of African languages publicly available.