Interspeech'2005 - Eurospeech
We compare the effectiveness of different syntactic features and representations for prosodic boundary prediction, setting out to clarify which representations are most suitable for this task. We took a machine learning approach, and ran a series of eight experiments. The results show that the representations have different strengths and that a combination yields the best result. We also find that linguistically deep representations do not yield clearly superior classifiers compared to classifiers obtained by extraction of shallow features.
Bibliographic reference. Ingulfsen, Tommy / Burrows, Tina / Buchholz, Sabine (2005): "Influence of syntax on prosodic boundary prediction", In INTERSPEECH-2005, 1817-1820.