Proceedings of the Workshop on Medical Speech Translation - MST '06

Abstract

Medical applications have emerged as one of the most popular domains for speech translation, and several functional systems now exist. Despite this, there is so far no established consensus on any of the central questions, including the following: •Does medical speech translation pose special problems, and if so, what are they? •What do the users (both doctors and patients) actually want? What constitutes acceptable performance, given that medicine is a safety-critical area? •What are the alternatives to speech translation for non-L1 speakers in healthcare situations? •What are the most important tasks, sub-domains and language pairs? •What architectures are most suitable for medical speech translation applications? (Fixed-phrase, ad hoc phrasal rules, rule-based, statistical...) •What evaluation/data collection methodologies are appropriate to medical speech translation? •What requirements are there on hardware platforms? What options currently exist? •How close are we to having applications that can be used in the field? In this one day workshop, our aim has been to get together as many as possible of the key players in this field, so that we can exchange information and clarify the above and other issues. We expect the workshop to be of interest to people working in all three component communities - speech technology, machine translation, and medicine. The main body of the workshop consists of two parts: oral presentation of papers, followed by a demo session. We will end with a panel discussion, which will include representatives of both the system developer and medical user communities.