This study aims to evaluate and compare the performance of automatic transcription systems for the ToBI (Tones and Breaks Indices) and ToDI (Transcription of Dutch Intonation) frameworks. Specifically, the focus is on matching or surpassing the results achieved by previous systems using a relatively small data set for training. By employing recent advancements in Natural Language Processing (NLP), this research demonstrates the potential to achieve comparable or superior performance in generating ToDI transcriptions of intonational phonology with limited labelled data available for boundary detection and boundary classification, while, for accent detection and accent classification, no results substantially better than the majority class baseline were obtained.

This paper was written as a BA thesis.