ROSETTA: Rhetorical and semantic environment for text alignment

In the framework of machine translation of multilingual parallel texts, the technique of alignment is based on statistical models and shallow linguistic parsing methods. When addressing the problem to corpora where different versions are derived or interpreted from the same source, we need further criteria to consider the forms of disparity between these versions. In this article we propose a content-driven approach based on the semantic and pragmatic structure of texts to aid in the process of alignment and comparison between their versions.