We build G2P models for Pashto, Tagalog and Lithuanian that significantly outperform a joint sequence model and a baseline recurrent neural network based model.

In this work, we devise new alignment strategies that work ef- fectively with recurrent neural network based models when only a small number of pronunciations ...

G2P models are particularly useful for low-resource languages that do not have well-developed pronunciation lexicons. Prominent G2P paradigms are based on ...

Traditional approaches for polyphone disambiguation are rule-based algorithms [3,4] and statistical machine learning methods [5,6].

With low-resource language pairs that do not have available and well-developed pronunciation lexicons, grapheme-to-phoneme models are particularly useful. These ...

These models are based on initial alignments be- tween grapheme source and phoneme target sequences. Inspired by sequence-to-sequence recurrent neural network– ...

Abstract. Transliteration is the process of converting a text in one script to another, guided by phonetic clues. This conversion requires an important set ...

This work proposes a G2P model based on a Long Short-Term Memory (LSTM) recurrent neural network (RNN) that has the flexibility of taking into consideration ...

Aug 4, 2017 · Low-resource grapheme-to-phoneme conversion us- ing recurrent neural networks. In Proc. ICASSP. Young-Bum Kim and Benjamin Snyder. 2012. Uni ...

There have been several approaches developed, including grapheme-to-phoneme conversion [20,21], based on statistics like machine translation [16, 22], as well ...