We argue that verbal patient diagnosis is a promising application for limited-domain speech translation, and describe an architecture designed for this type of task which represents a compromise between principled linguistics-based processing on the one hand and efficient phrasal translation on the other. We propose to demonstrate a prototype system instantiating this architecture, which has been built on top of the Open Source REGULUS 2 platform. The prototype translates spoken yes-no questions about headache symptoms from English to Japanese, using a vocabulary of about 200 words.