Moses does facilitate the construction of translation models from comparable data - but there has been some recent research on the topic that are a good starting point for developing such a tool:
On Sat, Oct 4, 2014 at 7:41 AM, javid dadashkarimi
<javiddadashkarimi at gmail.com> wrote:
> Is there any tool for extracting probabilistic bilingual dictionary for a
> bilingual comparable corpora? Does Moses support such a task?
> UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
> Corpora mailing list
> Corpora at uib.no