Accentuation of Adpositions and Particles in a Text-to-Speech System for Dutch

In this paper I propose an accent placement algorithm that locates accents on adpositions and particles for the use in a Dutch text-to-speech (TTS) system. The algorithm is intended to be a refinement of the rule that accents only content words, which is used in most TTS systems. Before the algorithm is set up, I discuss when adpositions and particles are accented in Dutch. For this empirical research, I made use of the Spoken Dutch Corpus (CGN) as empirical material. The combination of part-of-speech, syntactic as well as prosodic information for approximately 125,000 words in the CGN made it possible to determine whether the accentuation of adpositions and particles depends on their syntactic use within a sentence, or on the syntactic use of other constituents in the same sentence. The proposed accentuation algorithm takes a dependency tree with part-of-speech information as input.