Lecturer, Stellenbosch University
I am a lecturer in E&E engineering at Stellenbosch University, South Africa. Before moving back to South Africa, I did a postdoc at TTI-Chicago on multi-modal machine learning models combining speech and vision. I obtained my PhD from the University of Edinburgh, where I was supervised by Sharon Goldwater, Aren Jansen and Simon King; I worked on unsupervised speech processing, and played around with Bayesian and neural models.
My main interest is in the application of machine learning to problems in speech and language processing. I am particularly interested in methods that can learn from small amounts of labelled data, and in unsupervised methods that can learn directly from raw unlabelled data. Can an algorithm find meaningful units and structures in a corpus of speech audio, with only minimal guidance? How much supervision is required to build a useful speech system? These are central questions when building systems in low- and zero-resource settings.