Learning to Pronounce as Measuring Cross-Lingual Joint Orthography-Phonology Complexity

Abstract

Machine learning models allow us to compare languages by showing how hard a task in each language might be to learn and perform well on. Following this line of investigation, we explore what makes a language “hard to pronounce” by modelling the task of grapheme-to-phoneme (g2p) transliteration. By training a character-level transformer model on this task across 22 languages and measuring the model’s proficiency against its grapheme and phoneme inventories, we show that certain characteristics emerge that separate easier and harder languages with respect to learning to pronounce. Namely the complexity of a language's pronunciation from its orthography is due to the expressive or simplicity of its grapheme-tophoneme mapping. Further discussion illustrates how future studies should consider relative data sparsity per language to design fairer cross-lingual comparison tasks.

WITHDRAWN: Integrating Prior Knowledge and Data-Driven Approaches for Improving Grapheme-to-Phoneme Conversion in Korean Language

“…The reading and writing of Korean characters strictly adhere to the fundamental prin-ciple of left-to-right and top-to-bottom orientations. It is worth mentioning that the syllable structures in Korean can be classified as CVC, VC, CV, or V [24][25][26][27]. To provide a visual example of Korean text, Figure 1 illustrates a representative sample.…”