She often sends him voicemails in fon, a Beninese language, while he went to study in Russia. However, it does not include some of the phrases it uses.
“My mother can’t write fon fon and I don’t speak the language very well, but I speak French fluently,” Mr Dossou told the BBC.
“I often ask my sister to help me understand some of the phrases my mother uses,””, a-t-he said..
There is no question of improving one’s fon through study because, like hundreds of other African languages, it is mostly spoken and rarely documented, so that there are few, if any, books to teach grammar and syntax..
Driven by curiosity and fuelled by data collected on the fon through a Jehovah’s Witness bible, Mr. Dossou and Chris Emezue,a Nigerianfriend, developed a model of artificial intelligence (AI) language translation, similar to Google Translate, which they named FFR.
They are still working to improve their invention.
Both students are among several AI researchers who use African langues languages in natural language processing (NLP), an AI branch used to teach and help computers understand human langues languages.
If the world had not stopped in the wake of the Covid-19 pandemic, Mr. Dossou and Mr. Emezue would have presented their creation to Abeba hundreds of participants at one of the world’s s’est largest AI conferences, the ICLR, held this week in Addis Ababa,the Ethiopian capital. .
It was the first time the event was held in Africa.
Instead of cancelling the event,the organizers decided to hold it virtually.
AI innovations have been identified as the driving force behind en the “fourth industrial revolution” that will bring radical changes to almost every aspect of our lives, including the way we work.
Some analysts have called the big data that power AI systems new.
For now, Africa is seen as losing in its role in building the future of AI,as the majority of the continent’s estimated 2,000 languages are classified as “low resources”, meaning that there is a lack of data about them and/or that what is available has not been indexed and stored in formats that may be useful.
Bridging the language gap
African languages are not taken into account when creating ANP applications such as voice d’images assistants, image recognition software, traffic alert systems and others.
But African researchers are working to eliminate this handicap.
“We nous are working to put Africa on the en NLP and en AI research map,” Dr Ignatius Ezeaniof Lancaster University told the BBC.
“If you don’t make your language resources available to the public, free of charge and openly, researchers won’t have the data to find creative solutions. We will always have to depend on, say,Google to determine the direction of the search,” said Dr. Ezeani.
The conference in Ethiopia was to be an important event for African researchers who, among other challenges, were denied visas to attend ICLR conferences in the United States and Canada, keeping them away from global conversations about AI.
- Robots help kill coronaviruses
- Why your new co-worker could be a robot
- Robots, our future masons?
- These newcomers who are revolutionizing the space race
“The fact that the conference was not held in Addis was a blow, it would have resulted in a massive change in the diversity of the conference,” Jade Abbott, founder of Masakhane,a research movement for the automatic translation of African languages, told the langues BBC.
Masakhane,which means “We build together” in isiZulu, has 150 members in 20 African countries. Its membership is open to all those langues interested in language translation.
“We are building a community of people who care about African languages and who are willing to build translation models. 30% of the world’s languages are African, so why don’t we have30% of the NLP’s publications? PNL ? asked Mrs Abbott.
The network focuses on promoting language translation for Africans by Africans and encourages free sharing of resources and collaboration to help researchers build on the work of others..
However, most of the time it’s about starting from scratch.
A researcher affiliated with Masakhane, for example, is currently collecting data from speakers of Damara, a Khoisan language – famous for its rattling – in Namibia,Abbott said.
So far, Masakhane members have carried out 35 translations of 25 African languages, , a-t-she added.
en In addition to Masakhane, there are other initiatives to create and strengthen networks of AI researchers on the continent:
- Deep Learning Indaba, which promotes AI in Africa and organizes an annual conference
- Data Science Africa, which connects researchers on the continent
- BlackinAI, an initiative that promotes the inclusion of black people in the field of artificial intelligence
Dr Ezeani calls them the “silent struggles” of Africans working in the field of AI.
It considers that these commitments contribute to the development of the continent’s capabilities, both in terms of BUILDING AI infrastructure and the skills of researchers and developers.
“This is essential not only for recognition but also for addressing local challenges, for example in health, agriculture, education and governance,with local and targeted solutions,”he said.
“Perhaps we too can appropriate and control the discourse at some point,” donné”, a-t-he added.
Hey Alexa, are you talking about Igbo?
Dr. Ezeani is currently working on an automatic translation of the Igbo language from Nigeria to English.
“In five to ten years, I think I will be able to interact with Alexa in Igbo or in any language, which will be a huge feat and really satisfying,” said Dr. Ezeani.
Currently, none of the voice assistants of the major players in the global market, Alexa of Amazon,Siri of Apple and Google Home, do not operate in the African language. Google Translate is enabled for 13 langues African languages, including igbo, but it is far from perfect.
Dr Ezeani said the work he and others are doing could encourage technology companies to integrate African languages into their devices.
However, he cautions that African researchers working in the field of AI must be driven by original ideas “that are really useful to people” and not pursue futile projects.
“We can check whether,for example, exemple the translation from igbo to Yoruba and vice versa is actually more useful than igbo to English; or if speech or visual systems towards text are more necessary than text to text,””, a-t-he said.
As for Mr. Dossou and his co-inventor,inventeurMr. Emezue, they have great ambitions for the FFR if they can obtain funding.
They believe that fon, a Bantu language spoken by more than two million people in BéninBenin, as well as in parts of Nigeria and Togo, will help them develop their work in other markets.
Fon is part of the Niger-Congo language family, which means that it shares a common ancestral lineage, with languages spoken in parts of West, Central, Eastern and Southern Africa.
But for now, their goal is to continue training FFRs so they can better translate everyday conversations.
“Maybe inPeut-être about a year,000, my mother’s messages will be translated into French,” Mr. Dossou said.