Kambule Doctoral Award

The Kambule Doctoral Award, held in honour of one of South Africa’s greatest mathematician and teacher, Dr Thamsanqu W. Kambule, recognises  excellence in research and writing by doctoral candidates at African universities, in any area of computational and statistical sciences, and whose work demonstrates excellence through technical rigour, significance of contribution, and a strong role in strengthening African machine learning.  Its recipients are those that uphold Thamsanqa Kambule’s legacy as a defender of learning, a seeker of knowledge, and activist for equality.

The 2023 Kambule Doctoral Award has been given to Dr Arnu Pretorius, from Stellenbosch University, South Africa. 

Dr Pretorius received the Kambule Doctoral Award for his thesis titled ‘On noise regularised neural networks: initialisation, learning and inference’. When a neural learns from data, it can sometimes focus on irrelevant details that do not capture general aspects important for solving a particular task. This is referred to as “overfitting”. A powerful approach to address overfitting is to introduce noise into the training process. In this way, the network is forced to learn only from the more general signal that remains salient in the presence of noise, if it is to perform well. Arnu’s thesis focused on the mathematical underpinnings of how neural networks behave using this approach at the start (initialisation), middle (learning) and end of training (inference). The thesis provides a new theoretical understanding of the learning dynamics of training in this setting and uses the developed theory to derive novel ways of initialising neural networks trained with noise that improves performance during inference.

Dr Arnu Pretorius is a senior research scientist at InstaDeep, with a PhD in Computer Science from Stellenbosch University. Based in Cape Town, South Africa, Arnu is actively involved in cutting-edge research and development efforts related to multi-agent reinforcement learning, combinatorial optimisation, AI-driven drug discovery and AI for social impact and collaborates closely with applied teams to tackle real-world challenges at scale.  

Dr Arnu Pretorius with his team at InstaDeep, Cape Town, South Africa.

Runner Up Kambule Doctoral Award

This years’  Runner Up for the Kambule Doctoral Award went to Dr Oumaima Hourrane from Hassan II University of Casablanca, Morocco, for her thesis titled “Semantic Textual Similarity based on Deep Learning: Towards the Automatic Paraphrastic Detection of Monolingual and Cross-Lingual Plagiarism”. Dr Hourrane’s PhD thesis advances research on plagiarism detection across various aspects. The thesis introduced a range of methods to detect several plagiarism types, generating high-quality outcomes with focus on paraphrastic, translation, and idea-based plagiarism, and not literal plagiarism forms like copy-pasting. The focus is directed towards analyzing documents at the level of content, including citations and sentences. Her thesis significantly enhanced multiple plagiarism detection types such as extrinsic, intrinsic, and cross-lingual.

Dr. Oumaima Hourrane, an NLP researcher and lecturer, holds a PhD in Computer Science from Hassan II University of Casablanca and a Master’s degree in Computer Science from Cadi Ayyad University. Her research expertise lies in deep learning, focusing on advanced model design and implementation for various NLP tasks. Areas of specialisation include Multilingual Modelling, Semantic Textual Similarity, Representation Learning, GNNs for text, and Low-Resourced NLP. Alongside her academic pursuits, she actively engages in event organisation and volunteering, showcasing her commitment to knowledge advancement and community involvement.

Alele-Williams Masters Award

The Alele-Williams Masters Award, held in honour of Prof. Grace A. Alele-Williams, one of Nigeria’s most influential mathematicians and educators,  recognises excellence in research and writing by Masters candidates at African universities, in any area of computational and statistical sciences. This Award celebrates  those whose work demonstrates excellence through technical rigour, significance of contribution, and a strong role in strengthening African machine learning. Its recipients are those that uphold Grace Alele-Williams’ legacy as a defender of learning, a seeker of knowledge, and activist for equality.

Boago Okgetheng from the University of Botswana, Botswana, is this years’ Alele-Williams Masters Award recipient.

Boago received  the Alele-Williams Masters Award for his MSc thesis titled  “Tagging: Setswana Complex Qualificatives & Adverbs”, which advances Natural Language Processing (NLP) for Setswana. Setswana, a Bantu language spoken in several African countries presents a unique disjunctive writing style challenges of tagging complex parts of speech, particularly qualificatives, and adverbs formed by multiple words. Boago’s thesis addresses these challenges of tagging complex qualificatives and adverbs for Setswana.

Boago Okgetheng, born in Lotlhakane West, Botswana, embarked on an academic journey ignited by a profound fascination with language and technology. Following his undergraduate studies at the University of Botswana, he volunteered as a research assistant from 2016 to 2022 at the same university where he got introduced to the realm of NLP and registered to pursue an MSc in Computer Science in 2018 and graduated in 2020. Boago is now considering registering for a Ph.D. with a topic on NLP and machine learning, ultimately pushing the boundaries of linguistic technology.

Boago Okgetheng with his colleagues at the University of Botswana, Botswana.

Runner Up Alele-Williams Award

The Runner Up for the 2023 Alele-Williams  Award is  Masechaba Sydil Kupa from Rhodes University, South Africa, for her thesis titled “Semantic segmentation of Astronomical Radio Images: A Computer Vision Approach”. New types of radio telescopes, such as MeerKAT and ASKAP are generating data at a petabyte scale. Handling this massive volume of data using conventional methods has become a challenging task. In Sydil’s thesis, a novel approach was employed, leveraging deep learning techniques to detect and segment astronomical objects. This holds particular significance for the MeerKAT telescope and our  pursuit toward understanding the origins of our Universe.

Sydil Kupa hails from the northern region of South Africa, specifically a small locality named Lebowakgomo. Sydil holds a BSc in Actuarial Sciences from the University of Pretoria, South Africa, and a BSc Honours in Statistics from the University of the Western Cape, South Africa. In 2019, she embarked on her journey with the South African Radio Astronomy Observatory (SARAO) as a Graduate-in-Training. During this time, she worked as a Junior Software Engineer while simultaneously pursuing an MSc in Astrophysics at Rhodes University, South Africa. The latter endeavour was made possible through funding from SARAO. She successfully graduated with her MSc in March 2023. Presently, Sydil serves as a Software Engineer at SARAO, contributing to the development of software for the MeerKAT telescope.

Sydil Kupa with colleagues at the South African Radio Astronomy Observatory, Cape Town, South Africa.

Maathai Impact Award

The Maathai Impact Award is held in honour of Prof. Wangari Muta Maathai, Africa’s first female Nobel Laureate, globally for her contribution to democracy, peace, and sustainable development in Kenya as well as across the great African region. This Award recognises work by African innovators that shows impactful application of machine learning and artificial intelligence to positively impact Africa and her communities. This award reinforces the legacy of Wangari Maathai in acknowledging contributions – both intellectual as well as activism – at the intersection of technology and issues such as ecology, development, gender, and African cultures.  

This years’ Maathai Impact Award co-recipients are Zindi and VoteBot.  

Zindi hosts the largest community of African data scientists, working to solve the world’s most pressing challenges using machine learning and AI. Zindi  connects data scientists with organisations, and provide a place to learn, connect, and find a job. Zindi aims to transform the African continent and showcase African data science talent to the world.

Zindi was founded in 2018 with the goal of making AI more accessible to all. Over the last 5 years, Zindi has become a home for more than 65 000 data enthusiasts, students, engineers, analysts, and researchers who are looking to learn, connect and grow their skills in data science, AI and machine learning. As a data science competition platform, Zindi has given out more than $500 000 in prizes to users across 52 African countries and 185 countries worldwide. Zindi’s talented and passionate community has built cutting-edge machine learning models for companies and organisations like Google Deepmind, Fossil, Absa, Rand Merchant Bank, Uber, Lacuna Fund, CGIAR, and many many more. Zindi aims  to help every one of their community members find their dream job in data and AI, by connecting their community with the global demand for talent.

Picture of one of Zindi ambassadors and a group of Zindians at a local event in Tanzania.

VoteBot, co-recipient of the  2023 Maathai Impact Award, is an initiative by Justice Code Foundation, a CivicTech organization based in Zimbabwe. Justice Code Foundation’s mission is to leverage the power of technology for civic good. VoteBot is one of its initiatives. VoteBot is accessible on WhatsApp. WhatsApp accounts for more that 44% of internet usage in Zimbabwe, making VoteBot easily  accessible  in rural areas. A user wishing to use VoteBot just sends a “Hi” to a given number and starts interacting with the platform. VoteBot then provides  options to access a range of information such as voter registration centres across the country, a free ride (for young people) to go and register to vote, a link to view their voter registration status, and what the Constitution of Zimbabwe says about elections. Users can also use the platform to report any electoral malpractices in their local community and as well upload any evidence of electoral malpractices. 

VoteBot began in January 2022 ahead of the 2023 elections in Zimbabwe to enhance voter information and citizen observation of elections using Artificial Intelligence, targeting young people and women. VoteBot has worked with the African Union CivicTech Fund which saw user growth to +15 000 in 9 months. VoteBot has also partnered with Accountability Lab Zimbabwe to enhance electoral accountability in the 2023 elections.

Team members of Justice Code Foundation implementing the VoteBot project in Marange, a rural area in  Zimbabwe.