<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1309">
  <Title>Protein Name Tagging for Biomedical Annotation in Text</Title>
  <Section position="5" start_page="1" end_page="1" type="relat">
    <SectionTitle>
4 Related Work
</SectionTitle>
    <Paragraph position="0"> Tanabe and Wilbur (2002) use a hybrid approach of transformation-based learning (Brill Tagger) with rule-based post processing. An obvious drawback in their approach as with other rule-based approaches including Fukuda et al. (1998) is that the approaches cannot handle many correlated features. As pointed out in their paper, errors in the early stage of rule application are often propagated to the later stage, damaging the overall performance. In contrast, our method can deal with correlated features owing to the generalization characteristic of SVM.</Paragraph>
  </Section>
class="xml-element"></Paper>