Sequence-to-sequence neural network models for transliteration

TL;DR

This paper shows that neural sequence-to-sequence models achieve near state-of-the-art results in transliteration tasks and provides an open-source Arabic-English dataset and models to facilitate further research.

Contribution

It introduces a new Arabic-English transliteration dataset and trained models, making machine transliteration more accessible.

Findings

Neural sequence-to-sequence models achieve top performance in transliteration

Open-sourced dataset and models for Arabic-English transliteration

Results are close to the state of the art

Abstract

Transliteration is a key component of machine translation systems and software internationalization. This paper demonstrates that neural sequence-to-sequence models obtain state of the art or close to state of the art results on existing datasets. In an effort to make machine transliteration accessible, we open source a new Arabic to English transliteration dataset and our trained models.