College of Agriculture, Engineering
and Science (CAES)

Master’s Research Paves Way for IsiZulu Language Processing

Research into isiZulu language processing has earned Mr Sibonelo Dlamini a Master of Science degree in Computer Science.

Dlamini was supervised by Mr Edgar Jembere and Mr Anban Pillay.

Dlamini completed his schooling at George Campbell School of Technology in Durban after which he registered at UKZN for a degree in Electronic Engineering, soon realising, however, that he was more passionate about computer programming and switched to a BSc in Computer Sciences. ‘As I progressed through my undergraduate degree I fell more in love with the field of computer programming and have been ever since,’ he said.

After completing his BSc Honours degree in Computer Science, Dlamini progressed to masters level, focusing his research on Natural Language Processing. This field of study combines programming, machine learning and language – three areas Dlamini is passionate about. ‘It is exciting to do research in an area which is a key driver of the Fourth Industrial Revolution,’ he said.

In his research project, Dlamini dealt with isiZulu, which is an agglutinative language, meaning it has a complex internal structure which is constituted by numerous morphemes and thus makes it more difficult to develop language technologies for – unlike English which has a simpler internal structure. Dlamini tested the hypothesis that incorporating information about the internal structure of words of an agglutinative language would improve performance on the Word Sense Disambiguation (WSD) task. This is a machine-learning task which determines the correct sense of an ambiguous word in text.

Dlamini’s research goes a long way towards helping maintain the relevance of isiZulu in the current information age.  He explained that people stop using languages if they could not use technologies for tasks like voice recognition, information retrieval and grammar checking.  Through his research he hoped to develop state-of-the-art technologies for isiZulu, and thus preserve the Nguni language.

‘The motivation for my study was to stop the migration of language preference from isiZulu to English,’ he said. ‘It will preserve the invaluable cultural heritage that the language represents for a large section of South African society and retain access to indigenous knowledge which is encoded in the language.’

Dlamini explained that developing language technologies for a language was important because it maintained the relevance of the language in the information age. ‘If we don’t create space for indigenous African languages within this revolution, we may witness their rapid extinction and the erosion of African identity which will necessarily follow,’ he said. ‘Through my research, I hope to mitigate this potential hazard through the development of state-of-the-art technologies for agglutinative languages.’

Dlamini is currently continuing his research at PhD level, working towards creating an automatic speech recognition solution for isiZulu, ‘since speech is becoming an ubiquitous means of interfacing with computers.’ His future plans are to work full-time as an academic because of his passion in both research and teaching.

Dlamini’s supervisor, Mr Edgar Jembere, believes his work will serve as a benchmark for any future studies on WSD and isiZulu language processing.   ‘Sibonelo is a smart and hardworking student. I see him as a promising young Computer Scientist who will pursue a research career in Natural Language Processing,’ he said.

Some of Dlamini’s hobbies include reading, gym and jogging.

Words: Samantha Ngcongo

Photograph: Supplied