<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3239">
  <Title>A Boosting Algorithm for Classification of Semi-Structured Text</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> In this paper, we focused on an algorithm for the classification of semi-structured text in which a sentence is represented as a labeled ordered tree7. Our proposal consists of i) decision stumps that use subtrees as features and ii) Boosting algorithm in which the subtree-based decision stumps are applied as weak learners. Two experiments on opinion/modality classification tasks confirmed that sub-tree features are important.</Paragraph>
    <Paragraph position="1"> One natural extension is to adopt confidence rated predictions to the subtree-based weak learners. This extension is also found in BoosTexter and shows better performance than binary-valued learners.</Paragraph>
    <Paragraph position="2"> In our experiments, n-gram features showed comparable performance to dependency features. We would like to apply our method to other applications where instances are represented in a tree and their subtrees play an important role in classifications (e.g., parse re-ranking (Collins and Duffy, 2002) and information extraction).</Paragraph>
  </Section>
class="xml-element"></Paper>