Image copyright
Bonaventure Dossou
Image caption
Bonaventure Dossou is working on an AI model to help translate Fon to French
Bonaventure Dossou has been thinking a lot about how to improve phone conversations with his mother.
She often sends him voice messages in Fon, a Beninese language, as he is away studying in Russia. He, however, does not understand some of the phrases she uses.
“My mum cannot write Fon and I don’t speak the language very well but I’m fluent in French,” Mr Dossou told the BBC.
“I frequently ask my sister to help me understand some of the phrases mum uses,” he said.
Getty
Fon phrases in English
-
Nùkócé nɔn yìnMy name is
-
Oun yìn wàn nouwé I love you
-
Ouh fɔn gangjiI’m fine
-
NùnùɖùFood
Source: Bonaventure Dossou
Improving his Fon through study is out of the question because like hundreds of other African languages, it is mostly spoken and rarely documented, so there are few, if any, books to teach the grammar and syntax.
Driven by curiosity and powered by data scraped from a Fon to French Jehovah Witness Bible, Mr Dossou and Chris Emezue, a Nigerian friend, developed an Artificial Intelligence (AI) language translation model, similar to Google Translate, which they have named FFR. It is still a work in progress.
The two students are among several AI researchers using African languages in Natural Language Processing (NLP), a branch of AI used to teach and help computers understand human languages.
Had the world not ground to a halt following the Covid-19 pandemic, Mr Dossou and Mr Emezue would have presented their creation to hundreds of participants at one of the world’s biggest AI conferences, ICLR, in Ethiopia’s capital, Addis Ababa, this week.
It would have been the first time the event was held in Africa.
Instead of cancelling the event the organisers decided to hold it virtually.
You may also be interested in:
AI innovations have been singled out as the driver of the so-called fourth industrial revolution which will bring radical changes to almost every aspect of our lives including how we work.
Some analysts have called big data, which power AI systems, the new oil.
At the moment, Africa is seen as losing out in playing a role in shaping the AI future, because the majority of the continent’s estimated 2,000 languages are categorised as “low-resourced” meaning there’s a dearth of data about them and/or what is available has not been indexed and stored in formats that can be useful.
Fixing the languages gap
African languages are not considered when building NLP applications like voice assistants, image recognition software, traffic alerts systems and others.
But African researchers are working to eliminate this handicap.
“We are focused on placing Africa on the NLP and AI research map,” Dr Ignatius Ezeani, from the University of Lancaster, told the BBC.
“Unless you have your language resources publicly available, free and open, researchers will not have the data for creative solutions on the fly. We will always have to depend on, say, Google to determine the direction of research,” Dr Ezeani said.
Image copyright
Ignatius Ezeani
Image caption
Dr Ezeani is working on a machine translation project of Igbo to English
The conference in Ethiopia was set to be a big deal for African researchers who, among the other challenges they face, have been denied visas to attend past ICLR conferences held in the US and Canada, locking them out of global AI conversations.
“Not having the conference in Addis was a huge blow,