Laurette Marais, from the HLT group, received the Unisa Council Award for best Research Master’s Degree in the College of Science, Engineering and Technology for the academic year 2013. The award is given annually to top students in the College for a Master’s Degree.
Laurette’s research was based on Text-To-Speech (TTS), also known as speech synthesis, a tool which converts text into computer-generated speech. “One of the ongoing challenges in TTS is to improve the naturalness of a synthetic voice,” she says. For her Masters research, Laurette built the components of a system (for the English language) – which uses grammar – that is able to define the kinds of questions and answers the system must handle and a mark-up scheme to help a TTS engine put the emphasis on the correct words and phrases.
Laurette’s’ research used a specific theory about the “so-called” information structure of sentences to build a grammar which could understand the information structure of questions and generate appropriately structured answers. “A more traditional approach would have been to build a new voice with specially prepared data – and lots of it! With the grammar-based approach, however, I was able to use an existing synthetic voice and could manipulate its behaviour for my purposes,” she says.
“I was very grateful to receive recognition for hard work in a field that has the potential to make a great impact in South Africa,” says Laurette.
Laurette is currently enrolled as a PhD student in Computer Science at the University of Stellenbosch. “I hope to complete my PhD and go on to make a contribution to human language technology research in South Africa, especially for our resource scarce languages,” she says.