Emphasizing Syntax for French to German Machine Translation

This paper tackles the issue of automated translation from French to German, using syntactic analysis to enhance the results of lexical statistics approaches of these last years. It is a non symmetrical method aiming at producing correctly built sentences in the target language, from parsed sentences in source language. The idea is that translation between weakly divergent languages could be optimized relying on parsing output of the sentence to be translated. Then, after applying lexical transfer, the generation of the sentence in target language is performed through a set of light and recurrent transformations, applicable on the parsing tree structure. This paper describes a first study of divergence with an experiment on a corpus of aligned sentences extracted from the Little Prince of Saint-Exupéry.