Survey of Low-Resource Machine Translation

  • 2021-09-01 16:57:58
  • Barry Haddow, Rachel Bawden, Antonio Valerio Miceli Barone, JindÅ™ich Helcl, Alexandra Birch
  • 37

Abstract

We present a survey covering the state of the art in low-resource machinetranslation. There are currently around 7000 languages spoken in the world andalmost all language pairs lack significant resources for training machinetranslation models. There has been increasing interest in research addressingthe challenge of producing useful translation models when very littletranslated training data is available. We present a high level summary of thistopical field and provide an overview of best practices.

 

Quick Read (beta)

loading the full paper ...