<?xml version="1.0" standalone="yes"?>
<Paper uid="C00-1046">
  <Title>Automatic Refinement of a POS Tagger Using a Reliable Parser and Plain Text Corpora</Title>
  <Section position="8" start_page="318" end_page="318" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper we presented a new approach to acquiring linguistic knowledge automatically from plain text corpora, and proved its feasibility by the experiment. Our method utilizes well-developed modules in a NLP system, including a tagger and a parser, and enables us to extract wflid rules with high accuracy. It is robust in that the application of the extracted incorrect knowledge does not cause a serious performance deterioration.</Paragraph>
    <Paragraph position="1"> As a first step to obtaining lexically dependent knowledge, we examined the validity of obtained POS rules to measure the viability of our unsupervised learning method fi'om plain text corpora. In the future we will expand the size of training corpora and make use of invalid PA rules with a low adjustment ratio.</Paragraph>
  </Section>
class="xml-element"></Paper>