Automatic acquisition of transfer rules from translation examples

Abstract

In our research, we have developed a transfer-based machine translation architecture for the translation from Japanese into German. One main feature of the system is the fully automatic acquisition of transfer rules from translation examples by using structural matching between the parsing trees. The translation system has been implemented as part of a language learning environment with the aim to provide personalized translations for the students. In this paper we present our formalism to represent syntactic and transfer knowledge, and explain the various steps involved in acquiring and applying transfer rules.