<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1619">
  <Title>The Types and Distributions of Errors in a Wide Coverage Surface Realizer Evaluation</Title>
  <Section position="9" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> Recent empirical experiments on surface realizers have shown that grammars for generation can be effectively evaluated using large corpora. We have helped clarify to what extent errors in accuracy may be due to the corpora itself and in the transformation process necessary to convert annotated sentences into the surface realizer's notation. Furthermore, we have performed a set of quantitative, manual analyses that have classified with increasing rigor the types of syntactic phenomena missing from the generation grammar of FUF/SURGE . The results demonstrate that FUF/SURGE is surprisingly robust with coverage lacking for a few important but somewhat infrequent syntactic phenomena. Finally, we  have established a topline and baseline performance measure for use in future comparisons between surface realizers.</Paragraph>
  </Section>
class="xml-element"></Paper>