<?xml version="1.0" standalone="yes"?>
<Paper uid="P05-1041">
  <Title>High Precision Treebanking Blazing Useful Trees Using POS Information</Title>
  <Section position="9" start_page="336" end_page="336" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> We conducted an experiment to measure inter-annotator agreement for the Hinoki corpus. Three annotators marked up 5,000 sentences. Sentence agreement was an unparalleled 65.4%. The method used identi es problematic annotations as a byproduct, and allows the treebank to be improved as its underlying grammar improves. We also presented a method to speed up the annotation by exploiting existing part-of-speech tags. This led to a decrease in the number of annotation decisions of 19.5%.</Paragraph>
  </Section>
class="xml-element"></Paper>