This paper describes Sehda's S2MT (Syntactic Statistical Machine Translation) system submitted to the Korean-English track in the evaluation campaign of the IWSLT-05 workshop. The S2MT is a phrase-based statistical system trained on linguistically processed parallel data.