Learning English Syllabification Rules

摘要

This paper describes LE-SR (Learning English Syllabification Rules), the first machine learning program that learns English syllabi-fication rules, i.e., rules that tell how to divide English words into syllables for pronunication. LE-SR uses a unique knowledge representation called C-S-CL-SS which effectively generalizes English graphemes. Given a 20,000 on-line pronouncing dictionary, LE-SR learned 423 syllabification rules from 90