This paper presents a semi-automatic approach to translating the narrative descriptions of use cases to sequence diagrams. It presents rules for writing and rewriting use case descriptions that can be understood by both developers and domain experts and that can be translated mechanically to message sends, which are used to build sequence diagrams. The translation accommodates developer's instructions on how to translate some parts of a use case description while mechanically producing message records from the description. It deduces classes, objects, associations, attributes, and operations from use cases. A prototype of a use case translator that supports the approach is described here.