<?xml version="1.0" standalone="yes"?>
<Paper uid="H93-1040">
  <Title>EVALUATION OF MACHINE TRANSLATION</Title>
  <Section position="10" start_page="209" end_page="209" type="concl">
    <SectionTitle>
7. CONCLUSIONS
</SectionTitle>
    <Paragraph position="0"> The DARPA MT evaluation methodology strives to minimize the inherent subjectivity of judging translations, while optimizing the portability and replicability of test results and accommodating the variety of approaches, Languages, and end-user applications.</Paragraph>
    <Paragraph position="1"> The two evaluation suites described in this paper sought to accomplish these goals. The comparison among them accordingly is based upon the fidelity of the measurement, the efficiency of administration, and ultimately the portability of the test to other environments. We find, subject to further testing underway, that the second suite is advantageous in all these respects.</Paragraph>
  </Section>
class="xml-element"></Paper>