Part-of-Speech Tagging and Chunking in Text-to-Speech Synthesis for South African Languages

Text-to-speech synthesis can be an empowering communication tool in the hands of the print-disabled or augmentative and alternative communication user. In an effort to improve the naturalness of synthesised speech — and thus enhance the communication experience — we apply the natural language processing tasks of part-of-speech tagging and chunking to the text in the synthesis process. We cover the South African languages of (South African) English, Afrikaans, isiXhosa, isiZulu and Sepedi. The part-of-speech tagging delivers positive results for most of the languages; however, the chunking does not give any improvement in its current form.