The present work focuses on using tree-shaped syntactic structures as an intermediate sentence representation in an experimental English-Czech machine translation system.