<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-3802">
  <Title>Graph Based Semi-Supervised Approach for Information Extraction</Title>
  <Section position="3" start_page="9" end_page="9" type="intro">
    <SectionTitle>
2 Previous Work
</SectionTitle>
    <Paragraph position="0"> (Blum and Mitchell, 1998) proposed an approach based on co-training that uses unlabeled data in a particular setting. They exploit the fact that, for some problems, each example can be described by multiple representations. They develop a boosting scheme which exploits conditional independence between these representations.</Paragraph>
    <Paragraph position="1"> (Blum and Chawla, 2001) proposed a general approach utilizing unlabeled data by constructing a graph on all the data points based on distance relationships among examples, and then to use the known labels to perform a graph partitioning using the minimum cut that agrees with the labeled data.</Paragraph>
    <Paragraph position="2"> (Zhu et al., 2003) extended this approach by proposing a cut based on the assumption that labels are generated according to a Markov Random Field on the graph , (Joachims, 2003) presented an algorithm based on spectral graph partitioning.</Paragraph>
    <Paragraph position="3"> (Blum et al., 2004) extended the min-cut approach by adding randomness to the graph structure, their algorithm addresses several shortcomings of the basic mincut approach, yet it may not help in cases where the graph does not have small cuts for a given classification problem.</Paragraph>
  </Section>
class="xml-element"></Paper>