<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1002">
  <Title>Linear-Time Dependency Analysis for Japanese</Title>
  <Section position="11" start_page="4" end_page="4" type="concl">
    <SectionTitle>
7 Conclusion and Future Directions
</SectionTitle>
    <Paragraph position="0"> We have presented a novel algorithm for Japanese dependency analysis. The algorithm allows us to analyze dependency structures of a sentence in linear-time while keeping a state-of-the-art accuracy. We have shown a formal description of the algorithm and discussed it theoretically in terms of time complexity. In addition, we have evaluated its efficiency and performance empirically against the Kyoto University Corpus. Our parser gives the best accuracy, 89.56%, in the previously published results. null In the future, it would be interesting to apply this algorithm to speech recognition in which it is more desirable to analyze a sentence in a left-to-right manner. Another interesting direction would be to explore features for conjunctive structures. Although we found some useful features, they were not enough to improve the performance much. We expect stacking would be useful.</Paragraph>
  </Section>
class="xml-element"></Paper>