Retour au feed
arXiv cs.CL·

Assessing Dutch Syllabification Algorithms and Improving Accuracy by Combining Phonetic and Orthographic Information through Deep Learning

Signal
72
Hype
15
En 3 lignesÉtude comparative de quatre algorithmes de syllabification du néerlandais (Brandt Corstius, Liang, Trogkanis-Elkan CRF, et un modèle deep learning). Le modèle deep learning combinant informations phonétiques et orthographiques atteint 99,65% de précision (+0,14% vs littérature). Les algorithmes data-driven surpassent l'approche basée sur des règles.
Lire la source
Ton avis ?
PapersBenchmarksGénération de code

Résumé généré par Claude — vérifié par l'humain