Morphological Segmentation Can Improve Syllabification

Abstract

Syllabification is sometimes influenced by morphological boundaries. We show that incorporating morphological information can improve the accuracy of orthographic syllabification in English and German. Surprisingly, unsupervised segmenters, such as Morfessor, can be more useful for this purpose than the supervised ones.

On Pronunciations inWiktionary: Extraction and Experiments on Multilingual Syllabification and Stress Prediction

“…This highlights the contribution of other languages in a single massively multilingual model trained to do both tasks. Other researchers have found that good performance on syllabification requires much more data than this (Nicolai et al, 2016). We highlight the fact that many of the languages have less than 10 test examples and can be considered truly lowresource; the contribution of many other languages allows our multilingual models to predict the correct pronunciation with minimal training data in a specific language.…”

“…Weerasinghe et al, 2005;Müller, 2006) and data-driven approaches (e.g. Bartlett et al, 2008;Nicolai et al, 2016;Gyanendro Singh et al, 2016). However, previous work has focused primarily on a handful of languages, and some focus on orthographic syllabification rather than phonemic segmentation.…”