Hybrid Method for Stress Prediction Applied to GLAFF-IT, a Large-Scale Italian Lexicon

摘要

This paper presents a hybrid method for automatic stress prediction that we apply to GLAFF-IT, a large-scale Italian lexicon we extracted from GLAW-IT, a Machine-Readable Dictionary grounded on Wikizionario. Our approach combines heuristic rules and a logistic model trained on the words' sets of phonological features. This model reaches a 98.1% accuracy. The resulting resource is a large lexicon for the Italian language that we release under a free licence. It includes morphological and phonological information for each of its 457,702 entries. As of today, it is the only Italian lexicon featuring both large coverage and indication of stress position.