M.Sc., Faculty of Science, University of the Witwatersrand, 2011 / Studies have shown that text-to-speech systems need detailed prosodic models of a language in order to
ideally sound natural to native speakers of the language. A text-to-speech system developed for Sesotho
needs to have tone implemented in it since Sesotho is a tonal language which uses pitch variations to
distinguish lexical and/or grammatical meaning.
In order to implement tone for a language such as Sesotho, it is necessary for a tone modeling
algorithm to receive as input the tone labels of the syllables of a word. This allows the algorithm to
predict the appropriate intonation of the word. The aim of our study is to improve a basic tone labeling
algorithm that predicts tone labels using three Sesotho tonal rules. The application of this algorithm
is restricted to polysyllabic verb stems. The research study involves implementing an extended tone
labeling algorithm that implements four additional Sesotho tonal rules and extends its application to all
the other parts of speech.
The results of our study show that the extended tone labeling algorithm significantly improves the
basic algorithm by increasing the number of matched tone labels. Furthermore, our study provides the
basic step to tone modeling for languages such as Sesotho which do not mark tone labels in orthography.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:wits/oai:wiredspace.wits.ac.za:10539/11248 |
| Date | 06 February 2012 |
| Creators | Raborife, Mpho |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
Page generated in 0.0205 seconds