<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3708">
  <Title>S-MINDS 2-Way Speech-to-Speech Translation System</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Speech translation technology has the potential to give nurses and other clinicians immediate access to consistent, easy-to-use, and accurate medical interpretation for routine patient encounters. This could improve safety and quality of care for patients who speak a different language from that of the healthcare provider.</Paragraph>
    <Paragraph position="1"> The most common hospital interactions are interview-style dialogs where the provider's and patient's utterances are simple and relatively predictable. Sehda's speech translation system, S-MINDS, focuses on translating in such situations with extremely high accuracy.</Paragraph>
    <Paragraph position="2"> One key difference between S-MINDS and other speech translation systems is the amount of data required in development. Most other systems depend on a moderate amount of domain-specific data being available. If the data is not already available, it is extremely time- and labor-intensive for a developer to collect enough realistic data to effectively model a pure SMT system - even if the developer has direct access to a group of actual users for whom its system is being optimized.</Paragraph>
    <Paragraph position="3"> For this and other reasons, Sehda focuses on rapid building and deployment of speech translation systems for tasks or languages where little or no data is available.</Paragraph>
  </Section>
class="xml-element"></Paper>