Evaluation of a sentence accentuation algorithm for a dutch text-to-speech system

In this contribution an algorithm for the automatic assignment of sentence accents in written Dutch, developed by Kager and Quene, is evaluated experimentally. Results show that the output of the algorithm is judged significantly more adequate than accents randomly distributed over content words but significantly less adequate than the accents produced by a trained news broadcaster and those realized by t.he majority .of the subjects themselves. Insight was so.ught mto the basiS of the adequacy judgments and suggesuons are made for the improvement of the algorithm.