Low-Resource G2P and P2G Conversion with Synthetic Training Data
Bradley D. Hauer, Amir Ahmad Habibi, Yixing Luan, Arnob Mallik, Grzegorz Kondrak · 2020
This paper presents the University of Alberta systems and results in the SIGMOR-PHON 2020 Task 1: Multilingual Graphemeto-Phoneme Conversion.Following previous SIGMORPHON shared tasks, we define a lowresource setting with 100 training instances.We experiment with three transduction approaches in both standard and low-resource settings, as well as on the related task of phoneme-to-grapheme conversion.We propose a method for synthesizing training data using a combination of diverse models.