Human Language Technologies (HLT) - First Sepedi Speech Recognition system developed at the CSIR

"The research and development of Human Language Technologies (HLT) forms an integral part of addressing the significant need to empower all South Africa's citizens with information," said Barnard. "HLT allows people to interact with technology naturally, preventing the constraints of illiteracy, lack of education, language barriers and disability to hamper access to information. It also allows for information to be provided in non-traditional ways to technologically underserved regions of the country" he added.

Using an Open Source toolkit prepared by the ISTC and data recorded and prepared by the University of the North, the team spent an exhausting two days working through the entire process of developing an Hidden Markov Model-based automatic speech recognition system. The system utilises a Sepedi pronunciation dictionary, created by the University of the North. This dictionary was created using the DictionaryMaker, a new technology recently developed by the ISTC.

The initial Sepedi recognition system will form the basis for further research to be conducted at the University of the North, which will aim to improve the recognition accuracy of the initial baseline system. The ISTC and the University of the North will continue to collaborate on this project, with the aim to establish a solid body of knowledge with regard to the fast and effective development of speech recognition systems for South Africa's many languages.

The research team (from left to right): Tebogo Modiba, Marelie Davel, Etienne Barnard, Jonas Manamela and Kope Mamadisa		Kagiso Chikane providing some linguistic advice to the research team