Gloria Monica Tobechukwu Emezue, popularly known as GMT Emezue, is a Professor of English at the Alex Ekwueme Federal University, Ndufu Alike Ikwo in Ebonyi State, Nigeria.
She is one of the Principal Investigators (PI) for the NaijaVoices project. The scope of our project is to curate up to 500 hours of audio datasets in each of the three languages, namely Hausa, Igbo and Yoruba.
In this interview, GMT Emezue explained that the NaijaVoices dataset will assist in the development of impactful voice technologies in the health, education, economic, religious, agricultural sectors and all other areas of human life in Africa.
Excerpts:
Please, can you introduce yourselves?
I am Gloria Monica Tobechukwu Emezue, popularly known as GMT Emezue. I am a Professor of English at the Alex Ekwueme Federal University, Ndufu Alike Ikwo in Ebonyi State, Nigeria. I am a literary critic, but can best be described as an Afrocentric scholar with a passion for identifying and promoting positive African values, indigenous practices and knowledge systems,
which includes African languages.
I am one of the Principal Investigators (PI) for the NaijaVoices project. My Co-Principal Investigator is Chris Emezue.
Chris Emezue is a seasoned researcher affiliated with Mila-Quebec AI Institute in Canada, a prestigious research institute known for its groundbreaking scientific advancements in artificial intelligence.
As a researcher, Chris is committed to developing intelligent and reliable systems for under-represented contexts (this includes under-represented African and indigenous languages and cultures).
His research cuts across natural language processing, causality, and
reinforcement learning. As a dedicated contributor to the field of AfricaNLP, he has worked on several key projects to improve the representation of low-resource African language technologies and datasets (MMTAfrica, OkwuGbe, AfroDigits, IgboAPI).
Furthermore, as an entrepreneur, Chris is working at Lanfrica Labs, a research institute with the mission to foster
technological innovation in language technologies.
Can you tell us about the nature and scope of the project?
The title of our project is “NaijaVoices: Curation of Audio Datasets in the Hausa, Igbo, and Yoruba Languages.” These are the three major languages that are spoken in Nigeria. These three languages are estimated to be spoken by over 60 million people, both within and outside Nigeria and the African continent. The scope of our project is to curate up to 500 hours of audio
datasets in each of the three languages.
Thus, in all, we would have not less than 1,500 hours of audio datasets in these three languages. When some scholars saw our project proposal, we were told that we were being too ambitious because what we are proposing has never been achieved in any Nigerian language at such scale, as a result of the challenges that are strewn all over the path of its successful achievement.
But I am glad that we undertook it and as of today, we have successfully curated the 1,500 hours of audio datasets in these three Nigerian languages. So far, to the best of our knowledge, at the moment, the NaijaVoices project boasts the largest volume of audio datasets to be curated in any Nigerian language (and probably second largest in Africa, the largest being for Kinyarwanda by Digital Umuganda).
What will Africa as a Continent benefit from the project?
Across the globe, major speech recognition platforms like Amazon’s Alexa, Apple’s Siri, and Google’s Home dominate the market, yet they neglect the rich tapestry of African languages. Not a single native African language is supported by the current voice technologies widely used.
What we have succeeded in doing by creating huge hours of audio datasets in these three languages will be of immense benefit to many Africans in several ways. To start with, many African scholars, who had been interested in research involving voice technologies will now have available audio datasets that they can use in inventing bespoke voice technologies for Nigerian languages. In the past, we are aware that many companies and start-ups had
expressed interest in creating machine learning models or artificial intelligence
platforms/gadgets that could use African languages.
However, the major challenge had been
the non-availability of huge volumes of audio datasets in these African languages due to some challenges which included: the non-availability of millions of authentic scripted or non-scripted texts; non-availability of the sentence generating, voice recording, or validating platforms that have been tailored to handle the peculiarities of the African languages. For instance, in the area of gathering what we call authentic texts in African languages, our experience during this research has shown that most of what is littered on the internet are riddled with linguistic errors.
Our research has therefore been able to solve many of these challenges. For instance, we were able to devise new methods for creating authentic, contextually-grounded sentences, which we call the Sentence-Generating Method. Also, we observed that due to the peculiarity of some African languages, (tones, special characters, and so many other features) many of the already
existing audio-creating platforms do not support the peculiarities of African languages.
This led us to work with our partners in the Lanfrica Labs to create platforms such as the Lanfrica Voice Recorder and Lanfrica Sentence Generating Platform, which we utilized to scale our operations. These platforms were specifically tailored towards meeting most of the peculiarities of African languages.
Finally, a large population of Africans, who had earlier been excluded from many voice technologies due to age or literacy skills, can now benefit from the new inventions that will be created using these three African languages, because language cuts across all age groups,
genders, and boundaries.
In a nutshell, the NaijaVoices dataset will assist in the development of impactful voice technologies in the health sector, education sector, economic sector, religious sector, agricultural sector, and all other areas of human life in Africa.
Who are the key sponsors of the project and how were you able to emerge as one of its principal investigators?
Our financial sponsor is the Lacuna Fund which is a consortium of the following organizations: Canada’s International Development Research Centre (IDRC), Google.org, and The Rockefeller
Foundation. Our implementation partners include Lanfrica Labs, Global KANAC Ltd, Digital Learning Indaba (DLI), Masakhane, and a host of local communities.
I was able to emerge as a principal investigator as a result of my interest in Artificial Intelligence and my passion for promoting Afrocentric ideals and systems, of which African indigenous languages are part. When I started discussing some ideas with my co-principal investigator, we discovered that we shared some common interests concerning African languages, the Natural Language Processing (NLP) eco-system and artificial intelligence.
What actually led to the conceptualisation of the project?
Our major source of inspiration for the NaijaVoices project is the noticeable absence of Nigerian (African) indigenous languages in most AI technologies like Alexa, Siri, just to mention but a few. This is more critical in the area of voice technologies, which tend to reach many more people, including the very young (like toddlers), children, youth, old, blind, illiterate as well as the literate and educated. And these groups make up the larger population of Nigerians.
Our analysis of the problem promptly revealed that the major problem was the absence of needed quality audio datasets in these languages. The datasets enable the creation of AI technologies. So, we set out to bridge this gap. And the result is the NaijaVoices project.
What stage is the project as we speak?
We have successfully completed the NaijaVoices project. We have over 1,500 hours of audio datasets in the three major Nigerian languages. This comprises 1.3+ million original and
contextual sentences, and 1.9+ million audio clips. The resource is publicly accessible at https://naijavoices.com/.
Stepping down the project to Nigeria, what will Nigerians and various sectors of the economy benefit?
Nigerians of all ages and in all fields of life like education, sports, agriculture, and health just to mention but a few will benefit a lot from the voice technologies that will be built as a result of the
NaijaVoices resource our effort, which is like laying the AI foundation for AI developments in Nigeria.
Furthermore, we believe that since Nigeria is gifted with many talented young men and women who are interested in artificial intelligence and Machine Learning, they too can also benefit economically when they develop some technologies that people would find very useful.
What is the investment potentiality of the project?
The NaijaVoices Project has lots of investment potential for individuals and organizations that are into machine learning and artificial intelligence. What we have created has the possibility of
bringing forth many start-ups to emerge from Nigeria and outside the country. It has lots of investment opportunities, which we hope that many Nigerians could leverage to reverse their
economic fortunes and become wealth creators.
Does Africa as a Continent stand the chance to compete favorably with the Western World?
The African continent has a chance of competing favorably with the Western world when we talk about bringing forth good ideas. Many Africans have wonderful ideas about things that could be done to transform the continent. However, many of these ideas need financial sponsorship to become realities and it is in this area of financial sponsorship that Africa is lagging behind.
From the so-called African leaders occupying positions in government to rich individuals, it is the same story of the paucity of sponsorship for brilliant ideas that will develop the continent. Many Africans have yet to discover the will to create a benign economic environment and structure that would permit and assist more Africans to become wealthy, thus enabling more development within the continent.
In the area of data curation, machine learning, and artificial intelligence, the
West appears to have fared better because there are public-spirited individuals and governments, who understand the principle of sustainable development and economic growth.
A simple example is seen in the fact that the fund that enabled us to achieve the vision of the NaijaVoices project is from the West, even though we had earlier sent our proposal to the Nigerian government.
So, our problem in Africa is the lack of a large number of selfless and visionary citizens and leaders, who would sow financial seeds (by supporting brilliant ideas) for the sake of developing and sustaining Africa's tomorrow and the lives of future generations.
Post a Comment