Automatic induction of shallow-transfer rules for open-source machine translation

Abstract

This paper focuses on the inference of structural transfer rules for shallow-transfer machine translation (MT). Transfer rules are generated from alignment templates, like those used in statistical MT, that have been extracted from parallel corpora and extended with a set of restrictions that control their application. The experiments conducted using the open-source MT platform Apertium show that there is a clear improvement in translation quality as compared to word-for-word translation (when no transfer rules are used), and that the resulting translation quality is very close to the one obtained using hand-coded transfer rules. The method we present is entirely unsupervised and benefits from information in the rest of modules of the MT system in which the inferred rules are applied.