<?xml version="1.0" standalone="yes"?>
<Paper uid="C00-2110">
  <Title>Japanese Dependency Analysis using a Deterministic Finite State Transducer</Title>
  <Section position="8" start_page="3126" end_page="3126" type="concl">
    <SectionTitle>
7 Conclusion
</SectionTitle>
    <Paragraph position="0"> We proposed a Japmiese dependency analysis using a Deterministic Finite State Transducer (DFST). The system analyzes a sentence with fairly good accuracy of about 81% in 0.17 millisecond on average. It can be at)plied to a wide range of NLP applications, including real time Machine Translation, Information Ex- null traction and Speech Recognition. There are a number of eflbrts to improve the accuracy of Japanese dependency analysis (Haruno et.al, 1997) (Fujio, Matsumoto, 1998) (Kanaymna et.al, 2000). In particular, it is interesting to see that Kanwmna's method uses a similar idea, limiting the head candidates, in order to improve the accuracy. We are plmming to incorporate the fruit of these works to improve our accuracy. We believe our resem'ch is complementary to these research. We also believe the method proposed in this l)aper ln~\y be applic&gt; ble to other languages which share the chaiacteristics explained earlier, for example, Korean, ~lhrkish and others.</Paragraph>
  </Section>
class="xml-element"></Paper>