14th International Congress of Phonetic Sciences (ICPhS-14)

San Francisco, CA, USA
August 1-7, 1999

Pre-Processing Input Text: Improving Pronunciation for Dutch Text-to-Speech Systems

Arjan J. van Hessen (1), R. Jansen (2), Louis C. W. Pols (3)

(1) University of Twente, The Netherlands
(2) Comsys International BV, Zeist, The Netherlands
(3) Institute of Phonetic Sciences, University of Amsterdam, The Netherlands

To improve pronunciation of Dutch Text-To-Speech Synthesisers, a processor was built that tries to detect problematic cases in input texts and solve these automatically if possible. Its primary task is to realise pronounceable forms for numbers that do not have a straightforward pronunciation: structural and contextual information is used in an attempt to determine to what category a number belongs. Once categorised, each number is expanded according to the pronunciation conventions of its category. Moreover, pre-processing of raw ASCII text to a more readable text format is performed. It can be said that this pre-processor is a useful aid in improving performance at run-time, for example during the reading out aloud of emails, although ambiguity and redundancy in the input text illustrate the need for improved semantic and syntactic parsing to approach human text interpretation skills.

Full Paper

Bibliographic reference.  Hessen, Arjan J. van / Jansen, R. / Pols, Louis C. W. (1999): "Pre-processing input text: improving pronunciation for Dutch text-to-speech systems", In ICPhS-14, 2243-2246.