SACAIR Online Proceedings

SACAIR2022 Proceedings

The Proceedings of SACAIR2022 Online Conference, the 3rd Southern African Conference for Artificial Intelligence Research

The SACAIR Online Proceedings include all abstracts, as well as the full research papers accepted for presentation at the SACAIR2022 online conference (Vol.I, December 2022).

ISBN: 978-0-6397-1978-8 (e-book)

Please scroll down should you wish to download the individual papers.

Any enquiries can be made to sacairsc@gmail.com

The Proceedings of SACAIR2022 Online Conference, the 3rd Southern African Conference for Artificial Intelligence Research

The SACAIR Online Proceedings include all abstracts, as well as the full research papers accepted for presentation at the SACAIR2022 online conference (Vol.I, December 2022).

ISBN: 978-0-6397-1978-8 (e-book)

Any enquiries can be made to sacairsc@gmail.com

Copyright Notice

All rights reserved. No part of these proceedings may be reproduced, stored in a retrieval system, or transmitted, without prior written permission of the publisher.
The SACAIR 2022 organising committee is not responsible for errors or omissions from individual papers contained in these proceedings. Technical electronic anomalies are possible and unavoidable during the compilation process. The publisher is not responsible for the accuracy and validity of information contained in these papers, nor is it responsible for the final use to which this book may be used.
The SACAIR 2022 Proceedings Editors attest as follows: All conference paper submissions that appear in these proceedings have been through a double-blind peer review process prior to acceptance into the final conference programme.
Editors: Anban Pillay, Aurona Gerber, and Edgar Jembere.

Published Online by the SACAIR2022 Organising Committee
Private Bag X20
Hatfield
0028
ISBN: 978-0-6397-1978-8 (electronic)
© The Authors
December 2022

Preface

Message from the General Chairs

Dear authors and readers,

It is with great pleasure that we write this foreword to the proceedings of the third Southern African Conference for Artificial Intelligence Research (SACAIR 2022), held as a hybrid online and in-person event from the 5th to 9th December 2022. The program included an unconference for students on the 5th December 2022 (a student-driven event allowing students to interact with each other and with sponsors and potential employers), a day of tutorials on the 6th December, and the main conference from 7 − 9 December 2022.

SACAIR 2022 is the third international conference focused on Artificial Intelligence hosted by the SACAIR Steering Committee, an affiliate of the Centre for AI Research (CAIR), South Africa. The Centre for AI Research (CAIR)1 is a South African distributed research network established in 2011 that aims to build world-class Artificial Intelligence research capacity in South Africa. CAIR conducts foundational, directed, and applied research into various aspects of AI through its nine research groups based at six South African universities (the University of Pretoria, the University of KwaZulu-Natal, the University of Cape Town, Stellenbosch University, University of the Western Cape and North West University).

The inaugural CAIR conference, the Forum for AI Research (FAIR 2019) was held in Cape Town, South Africa, in December 2019, SACAIR 2020 was held in February 2021 after being postponed due to the Covid pandemic and SACAIR 2021 was an online event hosted by the University of KwaZulu-Natal, in Durban, in December 2021.

We are pleased that this, our third annual Southern African Conference for Artificial Intelligence Research (SACAIR), continued to enjoy the confidence of the South African artificial intelligence research community. The 2022 conference attracted support from both authors, who submitted high-quality research papers, as well as researchers who supported the conference by serving on the international program committee.

Any sufficiently advanced technology has the potential to transform society for better or worse. Artificial Intelligence technologies in general, and their current data-driven form, has the potential to transform our world for the better. However, especially in the context of machine learning applications, there are well founded concerns around fairness, structural bias and amplification of existing social stereotypes, privacy, transparency, accountability and responsibility, and trade-offs among all these concerns, especially within the context of security, robustness, and accuracy of AI systems. These issues talk directly to concerns around social justice that have become ever more important in the modern age. It was decided that the theme for SACAIR 2022 should be AI for Social Justice. The choice of conference theme was intended to ensure multi-disciplinary contributions that focus both on the technical aspects and social impact and consequences of AI technologies. To give expression to this, the conference was organized as a multi-track conference that would cover broad areas of Artificial Intelligence namely:

  • Algorithmic, Data Driven and Symbolic AI (Computer Science & Engineering)

  • Socio-technical and human-centered AI (Information Systems)

  • Responsible and Ethical AI (Philosophy and Law)

  • Inter- and transdisciplinary AI research

The accepted papers show a healthy balance between contributions from logic-based AI and those from data-driven AI, as the focus on knowledge representation and reasoning remains an important ingredient of studying and extending human intelligence. In addition, important contributions from the fields of social-technical and human-centred AI and responsible and ethical AI are reported in this volume.

We expect this multi- and interdisciplinary conference to grow into the premier AI conference in Southern Africa as it brings together nationally and internationally established and emerging researchers from across disciplines including Computer Science, Engineering, Mathematics, Statistics, Informatics, Philosophy and Law. The conference is also focused on cultivating and establishing a network of talented students working in AI from across Africa.

A conference of this nature is not possible without the hard work and contributions from many stakeholders. We extend our sincere gratitude to our sponsors: the Artificial Intelligence Journal (AIJ), the National Institute of Computational Sciences (NiTHeCS), the Centre for Artificial Intelligence Research (CAIR) and the BMW IT Hub South Africa. These sponsors have made it possible to offer generous scholarships to students and emerging academics to participate in the conference. We sincerely thank the technical chairs for their work in overseeing technical aspects of the conference and the publication of the two volumes of the proceedings, the international panel of reviewers, our keynotes, authors, and participants for their contributions. Finally, we extend our gratitude to the track chairs, the local organizing committee, student organizers, and the conference organizer for their substative contributions to the success of SACAIR 2022.

October 2022 Alta de Waal
Bruce Watson

Forward

Message from the Program Chairs

This online proceedings the Third Southern African Conference of Artificial Intelligence Research (SACAIR 2022) is presented in two volumes. The first is published as Volume 1734 of the Springer series, Communications in Computer and Information Science – CCIS (https://link.springer.com/book/10.1007/978-3-031-22321-1). This online proceedings contains the abstracts of these papers. Volume II presents the selected and revised papers accepted for presentation at the conference and published in full.

The inter- and trans-disciplinary nature of the SACAIR series of conferences in Artificial Intelligence is unique in providing a venue for researchers from a diverse set of disciplines that include Computer Science, Engineering, Information Systems, Law and Philosophy and the Humanities. The organization of such a conference has to carefully consider the differing research methods, interests, publication standards, and cultures of these disciplines. The conference was thus organized around four tracks: Algorithmic, Symbolic and Data-Driven AI (Computer Science and Engineering – CSE), Socio-technical and human-centered AI (Information Systems – IS), Responsible and Ethical AI (Philosophy and Law – PHIL) and Inter- and trans-disciplinary AI research.

The program committee comprised 112 members (representing some 43 research institutions), 28 of whom were from outside Southern Africa. Each paper was reviewed by at least two members of the program committee in a rigorous, double-blind process. Great care was taken to ensure the integrity of the conference including careful attention to avoid conflicts of interest. The following criteria were used to rate submissions and to guide decisions: relevance to SACAIR, significance, technical quality, scholarship, and presentation that included quality and clarity of writing.

We received just under 100 abstracts, and after submission and a first round of evaluation, 73 submissions were sent to our SACAIR program committee for review. The papers consisted of 54 in the CSE track, 11 in the IS track and 7 in the PHIL track. Twenty-six full research papers were selected for publication in the Springer CCIS volume (an acceptance rate of 35.6%), whilst a further 18 papers were accepted for inclusion in this online volume (24.7% acceptance rate). The total acceptance rate for publication in the two volumes was 60.2% for reviewed submissions. In total, four papers from the Responsible and Ethical AI track, eight papers from Socio-technical and human-centered AI track and 32 papers the Algorithmic, Symbolic and Data-Driven AI were accepted for publication in the two volumes.

Thank you to all the authors who submitted work of an exceptional standard to the conference and congratulations to the authors whose work was accepted for publication. We place on record our gratitude to the program committee members whose thoughtful and constructive comments were well received by the authors.

Papers in these two volumes are organised per the three tracks.

December 2022
Anban W. Pillay
Edgar Jembere
Aurona Gerber

Our Sponsors

Organization

General Chairs

Alta de WaalBMW IT Hub South Africa, South Africa &
 University of Pretoria, South Africa
Bruce WatsonStellenbosch University, South Africa

Technical Committee Chairs

Aurona GerberUniversity of Pretoria, South Africa
Edgar JembereUniversity of KwaZulu-Natal, South Africa
Anban PillayUniversity of KwaZulu-Natal, South Africa

Organizing Committee

Alta de WaalBMW IT Hub South Africa, South Africa &
 University of Pretoria, South Africa
Bruce WatsonStellenbosch University, South Africa
Aurona GerberUniversity of Pretoria, South Africa
Danie SmitBMW IT Hub South Africa, South Africa
Edgar JembereUniversity of KwaZulu-Natal, South Africa
Emile EngelbrechtStellenbosch University
Emma Ruttkamp-BloemUniversity of Pretoria, South Africa
Anban PillayUniversity of KwaZulu-Natal, South Africa

Program Committee

  1. Algorithmic, data-driven & symbolic AI track

    1. Track Chairs

        
      Alta de WaalBMW IT Hub South Africa, South Africa &
       University of Pretoria, South Africa
      Anban PillayUniversity of KwaZulu-Natal, South Africa
      Arina BritzStellenbosch University, South Africa
      Edgar JembereUniversity of KwaZulu-Natal, South Africa
      Ivan VarzinczakUniversity of Paris, France
      Terence Van ZylUniversity of Johannesburg, South Africa
    2. Program Committee

        
      Abiodun ModupeUniversity of Pretoria, South Africa
      Albert HelbergNorth-West University, South Africa
      Allan De FreitasUniversity of Pretoria, South Africa
      Andrew PaskaramoorthyUniversity of Cape Town, South Africa
      Anna Sergeevna BosmanUniversity of Pretoria, South Africa
      Bruce WatsonStellenbosch University, South Africa
      Bubacarr BahAfrican Institute for Mathematical Sciences (AIMS), South Africa
      Charis HarleyUniversity of Johannesburg, South Africa
      Colin ChibayaSol Plaatje University, South Africa
      David TomanUniversity of Waterloo, Canada
      Deon CotterrellUniversity of Johannesburg, South Africa
      Deshendran MoodleyUniversity of Cape Town, South Africa
      Duncan CoulterUniversity of Johannesburg, South Africa
      Dustin Van Der HaarUniversity of Johannesburg, South Africa
      Eduan KotzéUniversity of the Free State, South Africa
      Etienne BarnardNorth-West University, South Africa
      Fabio CozmanUniversity of São Paulo, Brazil
      Febe de WetStellenbosch University, South Africa
      Fred NicollsUniversity of Cape Town, South Africa
      Gift KhangamwaUniversity of the Witwatersrand, South Africa
      Giovanni CasiniISTI – CNR, Italy
      Guillermo R. SimariUniversidad del Sur in Bahia Blanca, Argentina
      Hairong WangUniversity of the Witwatersrand, South Africa
      Herman KamperStellenbosch University, South Africa
      Hima VadapalliUniversity of the Witwatersrand, South Africa
      Iliana M. PetrovaInria, France
      Inger Fabris-RotelliUniversity of Pretoria, South Africa
      Jaco VersfeldStellenbosch University, South Africa
      Jan BuysUniversity of Cape Town, South Africa
      Jesse HeyninckOpen Universiteit-the Netherlands, Netherlands
      Jiahao HuoUniversity of the Witwatersrand, South Africa
      Jules-Raymond TapamoUnivesity of KwaZulu-Natal, South Africa
      Justice EmuoyibofarheLadoke Akintola University of Technology, Nigeria
      Karim TabiaUniversité d’Artois, France
      Laura GiordanoDISIT, Università del Piemonte Orientale, Italy
      Laurent PerrusselIRIT – Universite de Toulouse, France
      Leopoldo BertossiSKEMA Business School Canada Inc., Canada
      Louise LeenenUniversity of the Western Cape, South Africa
      Makhamisa SenekaneUniversity of Johannesburg , South Africa
      Mandlenkosi GwetuUniversity of KwaZulu-Natal, South Africa
      Marelie DavelNorth-West University, South Africa
      Mohamed VariawaUniversity of Johannesburg, South Africa
      Olawande DaramolaCape Peninsula University of Technology, South Africa
      Paul AmayoUniversity of Cape Town, South Africa
      Pieter de VilliersUniversity of Pretoria, South Africa
      Ramon Pino PerezUniversité d’Artois, France
      Richard BoothCardiff University, Wales
      Richard KleinUniversity of the Witwatersrand, South Africa
      Roald EiselenNorth-West University, South Africa
      Rudzani MulaudziThe University of the Witwatersrand, South Africa
      Shakuntala BaichooUniversity of Mauritius, Mauritius
      Sihem BelabbesLIASD, Université Paris 8, France
      Stefan WoltranTU Wien, Austria
      Steven JamesUniversity of the Witwatersrand, South Africa
      Sunday OladejoStellenbosch University, South Africa
      Sunday OlatunjiImam Abdulrahman Bin Faisal University Dammam, Saudi Arabia
      Tevin MoodleyUniversity of Johannesburg, South Africa
      Thembinkosi SemwayoUniversity of the Witwatersrand, South Africa
      Thomas MeyerUniversity of Cape Town, South Africa
      Willie BrinkStellenbosch University, South Africa
      Zainoolabadien KarimUniversity of the Witwatersrand, South Africa
  2. Socio-technical and human-centered AI track

    1. Track Chairs

        
      Aurona GerberUniversity of Pretoria, South Africa
      Knut HinkelmannFHNW University of Applied Sciences and Arts Northwestern Switzerland, Switzerland
      Sunet EybersUniversity of Pretoria, South Africa
    2. Program Committee

        
      Andrea MartinFHNW University of Applied Sciences Northwestern Switzerland, Switzerland
      Catherine S. PriceUniversity of KwaZulu-Natal, South Africa
      Corné Van StadenUniversity of South Africa, South Africa
      Danie SmitBMW IT Hub South Africa, South Africa
      Douglas ParryStellenbosch University, South Africa
      Henk PretoriusUniversity of Pretoria, South Africa
      Johan BreytenbachUniversity of the Western Cape, South Africa
      Machdel MattheeUniversity of Pretoria, South Africa
      Marie HattinghUniversity of Pretoria, South Africa
      Patrick MikalefNorwegian University of Science and Technology (NTNU), Norway
      Phil van DeventerUniversity of Pretoria, South Africa
      Rennie NaidooUniversity of Pretoria, South Africa
      Riana SteynUniversity of Pretoria, South Africa
      Ridewaan HansloUniversity of Pretoria, South Africa
      Zola MahlazaUniversity of Pretoria, South Africa
  3. Responsible and ethical AI track

    1. Track Chairs

        
      Emma Ruttkamp-BloemUniversity of Pretoria, South Africa
      Tanya de Villiers-BothaStellenbosch University, South Africa
    2. Program Committee

        
      Andrea PalkStellenbosch University, South Africa
      Ann-Katrien OimannKU Leuven / Royal Military Academy, Belgium
      Arzu FormánekUniversity of Vienna, Austria
      Ashley CoatesUniversity of the Witwatersrand, South Africa
      Attlee Munyaradzi GamundaniNamibia University of Science and Technology, Namibia
      Cindy FriedmanUtrecht University, Netherlands
      Dilara BogaCentral European University (Vienna), Austria
      Fabio TollonStellenbosch University, South Africa
      Helen RobertsonUniversity of the Witwatersrand, South Africa
      Karabo MaiyaneUniversity of Pretoria, Nelson Mandela University, South Africa
      Mbangula Lameck AmugongoNamibia University of Science and Technology, Namibia
      Niël ConradieRWTH Aachen University, Germany
      Rosemann AchimDe Montfort University, United Kingdom
      Ryan NefdtUniversity of Cape Town, South Africa
      Sven NyholmEindhoven University of Technology, Netherlands
      Zach GudmunsenUniversity of Leeds, United Kingdom

SACAIR2022 Proceedings Volume I

Oosthuizen, A et. al

Adversarial Training for Channel State Information Estimation in LTE Multi-Antenna Systems-|- Andrew Oosthuizen, Albert Helberg and Marelie Davel

Hounkanrin, A et. al

Content-based medical image retrieval using a class similarity-aware cross-entropy loss-|- Anicet Hounkanrin, Paul Amayo and Fred Nicolls

Josias, S and Brink, W

Jacobian norm regularisation and conditioning in neural ODEs-|- Shane Josias and Willie Brink

Manaka, T et. al

Improving Cause-of-Death Classification from VerbalAutopsy Reports-|- Thokozile Manaka, Terence Van Zyl and Deepak Kar

Arendse L. J. et. al

Real Time In-Game Playstyle Classification Using A Hybrid Probabilistic Supervised Learning Approach-|- Lindsay John Arendse, Branden Ingram and Benjamin Rosman

Theunissen, W. et. al

The missing margin: How sample corruption affects distance to the boundary measurements in ANNs-|- Marthinus Wilhelmus Theunissen, Coenraad Mouton and Marelie Davel

Davidson, M. and Moodley, D.

Spatial-Temporal Graph Neural Networks For Weather Prediction in South Africa-|- Mikhail Davidson and Deshendran Moodley

Muthivhi, M. et. al

Multi-Modal Recommendation System with Auxiliary Information-|- Mufhumudzi Muthivhi, Terence van Zyl and Hairong Wang

Mlotshwa, T. et. al

Cauchy Loss Function: Robustness Under Gaussian and Cauchy Noise-|- Thamsanqa Mlotshwa, Heinrich van Deventer and Anna Sergeevna Bosman

Moodley, T. and Van Der Haar, D.

CASA: Cricket Action Similarity Assessment in video footage using deep metric learning-|- Tevin Moodley and Dustin Van Der Haar

Acton, S. and Buys, J.

From GNNs to Sparse Transformers: Graph-based architectures for Multi-hop Question Answering-|- Shane Acton and Jan Buys

Khangamwa, G. et. al

Towards a methodology for addressing missingness in datasets, with an application to demographic health datasets-|- Gift Khangamwa, Terence van Zyl and Clint van Alten

Wang, S. et. al

Defeasible Justification Using the KLM Framework-|- Steve Wang, Tommie Meyer and Deshen Moodley

Heyninck, J. and Meyer, T.

Relevance in the computation of non-monotonic inferences-|- Jesse Heyninck and Thomas Meyer

Asaju, C. and Vadapalli, H.

Adaptive Reasoning: An Affect Related Feedback Approach for Enhanced E-learning-|- Christine Asaju and Hima Vadapalli

Baas, M. et. al.

TransFusion: Transcribing Speech with Multinomial Diffusion-|- Matthew Baas, Kevin Eloff and Herman Kamper

Frost, G. et. al

Fine-tuned Self-Supervised Speech Representations for Language Diarization in Multilingual Code-Switched Speech-|- Geoffrey Frost, Emily Morris, Joshua Jansen van Vuren and Thomas Niesler

Dunn, J. and Suleman, H.

Evaluating Automated and Hybrid Neural Disambiguation for African Historical Named Entities-|- Jarryd Dunn and Hussein Suleman

Fourie, E. et. al

Neural speech recognition for whale call detection-|- Edrich Fourie, Marelie Davel and Jaco Versfeld

Sinclair, N. and Buys, J.

Self-Supervised Text Style Transfer with Rationale Prediction and Pretrained Transformers-|- Neil Sinclair and Jan Buys

Mabokela, K. R. and Schlippe, T.

AI for Social Good: Sentiment Analysis to Detect Social Challenges in South Africa-|- Koena Ronny Mabokela and TimSchlippe

Maeko, E. and van der Haar, D.

A Model for Biometric Selection in Public Services Sector-|- Elisa Maeko and Dustin van derHaar

Smit, D. et. al

Technology days: An AI democratisation journey begins with a single step-|- Danie Smit, Sunet Eybers, Alta de Waal and Nhlanhla Sibanyoni

Gerber, A. and Taljaard, T

The Preparation of South African Companies for theImpact of Artificial Intelligence-|- Aurona Gerber and Tiaan Taljaard

Tollen, F]]

Answerability, Accountability, and the Demands ofResponsibility-|- Fabio Tollon

Vorster, E

Does Counterfactual Reasoning Hold the Key toArtificial General Intelligence?-|- Ethan Vorster

SACAIR2022 Proceedings Volume II

Eiselen, R.

Afrikaans Text Embeddings for Sequence Labelling with Deep Neural Networks-|- Roald Eiselen

de Swardt, U. and Kamper, H.

How Machine Learning Can Aid South African Farmers’ Security: Unsupervised Livestock Trajectory Embeddings-|- Urs de Swardt and Herman Kamper

Ofosu, S. M. et. al

Learning to Pay Multiple Attention with Fully Convolutional Transformers-|- Samuel Ofosu Mensah, Bubacarr Bah and Willie Brink

Variawa, M. et. al

Exploring the effectiveness of surrogate-assisted evolutionary algorithms on the batch processing problem-|- Mohamed Variawa, Terence Van Zyl and Matthew Woolway

Davidson, N. and Booth, R.

Applying Route Optimisation to Fair Rota Generation for Home-Help Services-|- Naomi Davidson and Richard Booth

Mots’Oehli, M. and Chao Yang, Y.

Public Parking spot Detection And Geo-localization Using Transfer Learning-|- Moseli Mots’Oehli and Yao Chao Yang

Combrink, H. et. al

Comparing Synthetic Tabular Data Generation Between a Probabilistic Model and a Deep Learning Model for Education Use Cases-|- Herkulaas Combrink, Vukosi Marivate and Benjamin Rosman

Ncume, V. et. al

Volatility forecasting using Deep Learning and sentiment analysis-|- Vuyo Ncume, Terence van Zyl and Andrew Paskaramoorthy

Cohen, J. et. al

Model-based Defeasible Reasoning-|- Jaron Cohen, Carl Combrinck and Thomas Meyer

Rajcomar, S. et. al

Anomaly Detection in Continuous Stirred Reactor (CSTR) Using Deep Learning-|- Shikar Rajcomar, Edgar Jembere and Anban Pillay

Mathonsi, T. and Van Zyl, T.

Statistics and Deep Learning-based Hybrid Model for Interpretable Anomaly Detection-|- Thabang Mathonsi and Terence Van Zyl

Sathan, D. and Baichoo, S.

Breast Cancer Molecular subtyping using Deep learning and multi-omics dataset-|- Dassen Sathan and Shakuntala Baichoo

Ormond, E.

[Global To Local: South African Perspectives on AI Ethics Risks-|- Emile Ormond]

De Bruin, J. and Gerber, A.

AI Competencies for Competitive Advantage: A Systematic Literature Review-|- Jurgens De Bruin and Aurona Gerber

Breytenbach, J. et. al;

A process for embedding ethics into the Information Systems curriculum in South Africa-|- Johan Breytenbach, Yusuf Adams and Carolien van den Berg

Weber-Lewerenz, B.

CDR and Best Practices in AI in Construction – Catalysts and key to sustainability in the Digital Era-|- Bianca Weber-Lewerenz

Ruttkamp-Bloem, E.

AI Ethics as Reasoning-|- Emma Ruttkamp-Bloem

de Villiers-Botha, T.

Re-assessing Google as Epistemic Tool in the Age of Personalisation-|- Tanya de Villiers-Botha

Any enquiries can be made to sacairsc@gmail.com