Getting the best out of a mixed bag

This paper discusses the development and implementation of an approach to the combination of Rule Based Machine Translation, Statistical Machine Translation and Translation Memory tecnologies. The machine translation system itself draws upon translation memories and both syntactically and statistically generated phrase tables, unresolved sentences being fed to a Rules Engine. The output of the process is a TMX file containing a varying mixture of TMgenerated and MT-generated sentences. The author has designed this workflow using his own language engineering tools written in Java.